fix(channels): auto-reclaim orphaned bot installations + accurate rebind conflict copy

channel_installation has no FK to workspace/agent (MUL-3515 §4), so deleting a workspace or hard-deleting an agent left the row behind, occupying the (channel_type, app_id) routing slot forever — the bot could never be rebound and the UI had no way to clear it (#4810). The 409 also always blamed "a different Multica workspace" even when the real owner sat in the same workspace. Auto-reclaim on delete: - DeleteWorkspace and the runtime-teardown paths now sweep the workspace's / archived agents' channel installations and every dependent row in-tx. - The shared install path (Feishu + Slack) reclaims a DEAD prior owner — a revoked placeholder or an orphan whose workspace/agent is gone — before the upsert, healing installations stranded before this fix. A live owner (active agent, including an archived one) is left in place, not stolen. Accurate conflict copy: - A rebind refused by a LIVE owner now distinguishes same-workspace / another agent, an archived agent, and a genuinely different workspace, for both Slack (typed sentinels) and Feishu (registration message). MUL-3937 Co-authored-by: multica-agent <github@multica.ai>
2026-08-02 18:13:27 +02:00 · 2026-07-08 19:31:55 +08:00
parent c05b67ae4a
commit b4aeb56e94
15 changed files with 957 additions and 195 deletions
--- a/server/internal/handler/runtime.go
+++ b/server/internal/handler/runtime.go
@@ -760,6 +760,14 @@ func (h *Handler) DeleteAgentRuntime(w http.ResponseWriter, r *http.Request) {
 		writeError(w, http.StatusInternalServerError, "failed to clean up agent invocation targets")
 		return
 	}
+	// Same app-layer cleanup for channel installations: channel_* has no
+	// workspace/agent FK (MUL-3515 §4), so an archived agent's bot installations
+	// would otherwise survive the hard-delete as orphans and keep occupying their
+	// (channel_type, app_id) routing slots, making those bots un-rebindable (#4810).
+	if err := qtx.DeleteChannelInstallationsByArchivedRuntimeAgents(r.Context(), rt.ID); err != nil {
+		writeError(w, http.StatusInternalServerError, "failed to clean up channel installations")
+		return
+	}
 	if err := qtx.DeleteArchivedAgentsByRuntime(r.Context(), rt.ID); err != nil {
 		writeError(w, http.StatusInternalServerError, "failed to clean up archived agents")
 		return
@@ -1004,6 +1012,14 @@ func (h *Handler) ArchiveAgentsAndDeleteRuntime(w http.ResponseWriter, r *http.R
 		writeError(w, http.StatusInternalServerError, "failed to clean up agent invocation targets")
 		return
 	}
+	// Same app-layer cleanup for channel installations: channel_* has no
+	// workspace/agent FK (MUL-3515 §4), so an archived agent's bot installations
+	// would otherwise survive the hard-delete as orphans and keep occupying their
+	// (channel_type, app_id) routing slots, making those bots un-rebindable (#4810).
+	if err := qtx.DeleteChannelInstallationsByArchivedRuntimeAgents(r.Context(), rt.ID); err != nil {
+		writeError(w, http.StatusInternalServerError, "failed to clean up channel installations")
+		return
+	}
 	if err := qtx.DeleteArchivedAgentsByRuntime(r.Context(), rt.ID); err != nil {
 		writeError(w, http.StatusInternalServerError, "failed to clean up archived agents")
 		return
--- a/server/internal/handler/runtime_profile.go
+++ b/server/internal/handler/runtime_profile.go
@@ -458,6 +458,15 @@ func (h *Handler) DeleteRuntimeProfile(w http.ResponseWriter, r *http.Request) {
 			writeError(w, http.StatusInternalServerError, "failed to clean up agent invocation targets")
 			return
 		}
+		// Same app-layer cleanup for channel installations: channel_* has no
+		// workspace/agent FK (MUL-3515 §4), so an archived agent's bot
+		// installations would otherwise survive the hard-delete as orphans and
+		// keep occupying their (channel_type, app_id) routing slots, making those
+		// bots un-rebindable (#4810).
+		if err := qtx.DeleteChannelInstallationsByArchivedRuntimeAgents(r.Context(), rid); err != nil {
+			writeError(w, http.StatusInternalServerError, "failed to clean up channel installations")
+			return
+		}
 		if err := qtx.DeleteArchivedAgentsByRuntime(r.Context(), rid); err != nil {
 			writeError(w, http.StatusInternalServerError, "failed to clean up archived agents")
 			return
--- a/server/internal/handler/slack.go
+++ b/server/internal/handler/slack.go
@@ -145,8 +145,12 @@ func (h *Handler) RegisterSlackBYO(w http.ResponseWriter, r *http.Request) {
 		switch {
 		case errors.Is(err, slack.ErrInvalidBotToken), errors.Is(err, slack.ErrInvalidAppToken), errors.Is(err, slack.ErrTokenAppMismatch):
 			writeError(w, http.StatusBadRequest, err.Error())
+		case errors.Is(err, slack.ErrTeamOwnedBySameWorkspace):
+			writeError(w, http.StatusConflict, "this Slack app is already connected to another agent in this workspace — disconnect it there first, then connect it here")
+		case errors.Is(err, slack.ErrTeamOwnedByArchivedAgent):
+			writeError(w, http.StatusConflict, "this Slack app is connected to an archived agent in this workspace — restore that agent, or disconnect its bot, before connecting it here")
 		case errors.Is(err, slack.ErrTeamOwnedByAnotherWorkspace):
-			writeError(w, http.StatusConflict, "this Slack app is already connected to a different Multica workspace")
+			writeError(w, http.StatusConflict, "this Slack app is already connected to a different Multica workspace — disconnect it there before connecting it here")
 		default:
 			// The dominant non-sentinel failure here is auth.test rejecting the
 			// pasted bot token (a user error), so guide the user to recheck the
--- a/server/internal/integrations/lark/channel_cleanup_test.go
+++ b/server/internal/integrations/lark/channel_cleanup_test.go
@@ -0,0 +1,201 @@
+package lark
+
+import (
+	"context"
+	"testing"
+
+	"github.com/jackc/pgx/v5/pgxpool"
+
+	"github.com/multica-ai/multica/server/internal/util"
+	db "github.com/multica-ai/multica/server/pkg/db/generated"
+)
+
+// Delete-time channel cleanup fixtures. These pin the #4810 fix on the OTHER
+// half of "auto-reclaim on delete": deleting a workspace, or hard-deleting an
+// archived agent on runtime teardown, must sweep the channel_installation rows
+// (and every dependent row) their owners left behind — channel_* has no FK to
+// workspace/agent (MUL-3515 §4), so nothing else would.
+const (
+	ccWS         = "cc000000-0000-4000-8000-000000000001"
+	ccRuntime    = "cc000000-0000-4000-8000-000000000003"
+	ccAgentArch  = "cc000000-0000-4000-8000-00000000000a"
+	ccAgentLive  = "cc000000-0000-4000-8000-00000000000b"
+	ccInstaller  = "cc000000-0000-4000-8000-000000000005"
+	ccUser       = "cc000000-0000-4000-8000-000000000006"
+	ccChatArch   = "cc000000-0000-4000-8000-0000000000c1"
+	ccChatLive   = "cc000000-0000-4000-8000-0000000000c2"
+	ccTokenArch  = "cc_token_archived"
+	ccTokenLive  = "cc_token_live"
+	ccAuditArch  = "ev_cc_archived"
+	ccAuditLive  = "ev_cc_live"
+	ccAppArchive = "cli_cc_archived"
+	ccAppLive    = "cli_cc_live"
+
+	// Workspace-delete fixture (its own workspace so the DeleteWorkspace can
+	// remove it, cascading the runtime + agent).
+	ccWSDel      = "cc000000-0000-4000-8000-000000000002"
+	ccRuntimeDel = "cc000000-0000-4000-8000-000000000004"
+	ccAgentDel   = "cc000000-0000-4000-8000-00000000000d"
+	ccChatWs     = "cc000000-0000-4000-8000-0000000000c3"
+	ccTokenWs    = "cc_token_ws"
+	ccAuditWs    = "ev_cc_ws"
+	ccAppWs      = "cli_cc_ws"
+)
+
+// seedFullInstallation inserts one active installation plus the full spread of
+// dependents (member link, chat-session binding, pending binding token, inbound
+// audit) so a cleanup path can be shown to sweep all of them.
+func seedFullInstallation(t *testing.T, ctx context.Context, pool *pgxpool.Pool, ws, agent, app, chatSess, tokenHash, auditEvent string) string {
+	t.Helper()
+	var id string
+	if err := pool.QueryRow(ctx, `
+INSERT INTO channel_installation (workspace_id, agent_id, channel_type, config, installer_user_id, status)
+VALUES ($1, $2, 'feishu', jsonb_build_object('app_id', $3::text), $4, 'active')
+RETURNING id
+`, ws, agent, app, ccInstaller).Scan(&id); err != nil {
+		t.Fatalf("seed installation app=%s: %v", app, err)
+	}
+	exec := func(q string, args ...any) {
+		if _, err := pool.Exec(ctx, q, args...); err != nil {
+			t.Fatalf("seed dependent for app=%s: %v", app, err)
+		}
+	}
+	exec(`INSERT INTO channel_user_binding (workspace_id, multica_user_id, installation_id, channel_type, channel_user_id)
+VALUES ($1, $2, $3, 'feishu', 'ou_cc_user')`, ws, ccUser, id)
+	exec(`INSERT INTO channel_chat_session_binding (chat_session_id, installation_id, channel_type, channel_chat_id, chat_type)
+VALUES ($1, $2, 'feishu', 'oc_cc_chat', 'p2p')`, chatSess, id)
+	exec(`INSERT INTO channel_binding_token (token_hash, workspace_id, installation_id, channel_type, channel_user_id, expires_at)
+VALUES ($1, $2, $3, 'feishu', 'ou_cc_user', now() + interval '10 minutes')`, tokenHash, ws, id)
+	exec(`INSERT INTO channel_inbound_audit (installation_id, channel_type, event_type, channel_event_id, drop_reason)
+VALUES ($1, 'feishu', 'im.message.receive_v1', $2, 'test')`, id, auditEvent)
+	return id
+}
+
+func ccCount(t *testing.T, ctx context.Context, pool *pgxpool.Pool, q string, args ...any) int {
+	t.Helper()
+	var n int
+	if err := pool.QueryRow(ctx, q, args...).Scan(&n); err != nil {
+		t.Fatalf("count: %v", err)
+	}
+	return n
+}
+
+// assertInstallationSwept checks an installation and all its dependents are gone,
+// with the inbound-audit row preserved but detached (installation_id NULL).
+func assertInstallationSwept(t *testing.T, ctx context.Context, pool *pgxpool.Pool, id, auditEvent string) {
+	t.Helper()
+	if n := ccCount(t, ctx, pool, `SELECT count(*) FROM channel_installation WHERE id = $1`, id); n != 0 {
+		t.Fatalf("installation not swept: %d rows remain (its bot's app_id slot stays occupied)", n)
+	}
+	if n := ccCount(t, ctx, pool, `SELECT count(*) FROM channel_user_binding WHERE installation_id = $1`, id); n != 0 {
+		t.Fatalf("member links not swept: %d dangling rows", n)
+	}
+	if n := ccCount(t, ctx, pool, `SELECT count(*) FROM channel_chat_session_binding WHERE installation_id = $1`, id); n != 0 {
+		t.Fatalf("chat-session bindings not swept: %d dangling rows", n)
+	}
+	if n := ccCount(t, ctx, pool, `SELECT count(*) FROM channel_binding_token WHERE installation_id = $1`, id); n != 0 {
+		t.Fatalf("binding tokens not swept: %d dangling rows", n)
+	}
+	if n := ccCount(t, ctx, pool, `SELECT count(*) FROM channel_inbound_audit WHERE channel_event_id = $1 AND installation_id IS NULL`, auditEvent); n != 1 {
+		t.Fatalf("audit row should survive detached (installation_id NULL), got %d", n)
+	}
+}
+
+func assertInstallationIntact(t *testing.T, ctx context.Context, pool *pgxpool.Pool, id, auditEvent string) {
+	t.Helper()
+	if n := ccCount(t, ctx, pool, `SELECT count(*) FROM channel_installation WHERE id = $1`, id); n != 1 {
+		t.Fatalf("a live-owner installation was swept: %d rows (want 1)", n)
+	}
+	if n := ccCount(t, ctx, pool, `SELECT count(*) FROM channel_user_binding WHERE installation_id = $1`, id); n != 1 {
+		t.Fatalf("live-owner member link was swept: %d (want 1)", n)
+	}
+	if n := ccCount(t, ctx, pool, `SELECT count(*) FROM channel_inbound_audit WHERE channel_event_id = $1 AND installation_id = $2`, auditEvent, id); n != 1 {
+		t.Fatalf("live-owner audit ref was detached: %d (want 1)", n)
+	}
+}
+
+// TestDeleteChannelInstallationsByArchivedRuntimeAgents: runtime teardown hard-
+// deletes archived agents; this cleanup must sweep exactly those agents'
+// installations (and dependents), leaving live agents on the runtime untouched.
+func TestDeleteChannelInstallationsByArchivedRuntimeAgents(t *testing.T) {
+	pool := channelScopeTestDB(t)
+	ctx := context.Background()
+	q := db.New(pool)
+
+	clean := func() {
+		_, _ = pool.Exec(ctx, `DELETE FROM channel_installation WHERE config->>'app_id' = ANY($1)`, []string{ccAppArchive, ccAppLive})
+		_, _ = pool.Exec(ctx, `DELETE FROM channel_user_binding WHERE multica_user_id = $1`, ccUser)
+		_, _ = pool.Exec(ctx, `DELETE FROM channel_chat_session_binding WHERE chat_session_id = ANY($1)`, []string{ccChatArch, ccChatLive})
+		_, _ = pool.Exec(ctx, `DELETE FROM channel_binding_token WHERE token_hash = ANY($1)`, []string{ccTokenArch, ccTokenLive})
+		_, _ = pool.Exec(ctx, `DELETE FROM channel_inbound_audit WHERE channel_event_id = ANY($1)`, []string{ccAuditArch, ccAuditLive})
+		_, _ = pool.Exec(ctx, `DELETE FROM workspace WHERE id = $1`, ccWS)
+	}
+	clean()
+	t.Cleanup(clean)
+
+	exec := func(query string, args ...any) {
+		if _, err := pool.Exec(ctx, query, args...); err != nil {
+			t.Fatalf("seed owner: %v", err)
+		}
+	}
+	exec(`INSERT INTO workspace (id, name, slug, description) VALUES ($1, 'cc ws', 'cc-ws', '')`, ccWS)
+	exec(`INSERT INTO agent_runtime (id, workspace_id, name, runtime_mode, provider)
+VALUES ($1, $2, 'cc runtime', 'local', 'multica_daemon')`, ccRuntime, ccWS)
+	exec(`INSERT INTO agent (id, workspace_id, name, runtime_mode, runtime_id, archived_at)
+VALUES ($1, $2, 'cc archived agent', 'local', $3, now())`, ccAgentArch, ccWS, ccRuntime)
+	exec(`INSERT INTO agent (id, workspace_id, name, runtime_mode, runtime_id)
+VALUES ($1, $2, 'cc live agent', 'local', $3)`, ccAgentLive, ccWS, ccRuntime)
+
+	archivedID := seedFullInstallation(t, ctx, pool, ccWS, ccAgentArch, ccAppArchive, ccChatArch, ccTokenArch, ccAuditArch)
+	liveID := seedFullInstallation(t, ctx, pool, ccWS, ccAgentLive, ccAppLive, ccChatLive, ccTokenLive, ccAuditLive)
+
+	if err := q.DeleteChannelInstallationsByArchivedRuntimeAgents(ctx, util.MustParseUUID(ccRuntime)); err != nil {
+		t.Fatalf("DeleteChannelInstallationsByArchivedRuntimeAgents: %v", err)
+	}
+
+	assertInstallationSwept(t, ctx, pool, archivedID, ccAuditArch)
+	assertInstallationIntact(t, ctx, pool, liveID, ccAuditLive)
+}
+
+// TestDeleteWorkspace_SweepsChannelInstallations: deleting a workspace must sweep
+// its channel installations (and dependents) so no orphan keeps occupying its
+// bot's routing slot after the workspace — and its cascade-deleted agents — are
+// gone.
+func TestDeleteWorkspace_SweepsChannelInstallations(t *testing.T) {
+	pool := channelScopeTestDB(t)
+	ctx := context.Background()
+	q := db.New(pool)
+
+	clean := func() {
+		_, _ = pool.Exec(ctx, `DELETE FROM channel_installation WHERE config->>'app_id' = $1`, ccAppWs)
+		_, _ = pool.Exec(ctx, `DELETE FROM channel_user_binding WHERE multica_user_id = $1`, ccUser)
+		_, _ = pool.Exec(ctx, `DELETE FROM channel_chat_session_binding WHERE chat_session_id = $1`, ccChatWs)
+		_, _ = pool.Exec(ctx, `DELETE FROM channel_binding_token WHERE token_hash = $1`, ccTokenWs)
+		_, _ = pool.Exec(ctx, `DELETE FROM channel_inbound_audit WHERE channel_event_id = $1`, ccAuditWs)
+		_, _ = pool.Exec(ctx, `DELETE FROM workspace WHERE id = $1`, ccWSDel)
+	}
+	clean()
+	t.Cleanup(clean)
+
+	exec := func(query string, args ...any) {
+		if _, err := pool.Exec(ctx, query, args...); err != nil {
+			t.Fatalf("seed owner: %v", err)
+		}
+	}
+	exec(`INSERT INTO workspace (id, name, slug, description) VALUES ($1, 'cc ws del', 'cc-ws-del', '')`, ccWSDel)
+	exec(`INSERT INTO agent_runtime (id, workspace_id, name, runtime_mode, provider)
+VALUES ($1, $2, 'cc runtime del', 'local', 'multica_daemon')`, ccRuntimeDel, ccWSDel)
+	exec(`INSERT INTO agent (id, workspace_id, name, runtime_mode, runtime_id)
+VALUES ($1, $2, 'cc agent del', 'local', $3)`, ccAgentDel, ccWSDel, ccRuntimeDel)
+
+	id := seedFullInstallation(t, ctx, pool, ccWSDel, ccAgentDel, ccAppWs, ccChatWs, ccTokenWs, ccAuditWs)
+
+	if err := q.DeleteWorkspace(ctx, util.MustParseUUID(ccWSDel)); err != nil {
+		t.Fatalf("DeleteWorkspace: %v", err)
+	}
+
+	if n := ccCount(t, ctx, pool, `SELECT count(*) FROM workspace WHERE id = $1`, ccWSDel); n != 0 {
+		t.Fatalf("workspace not deleted: %d rows", n)
+	}
+	assertInstallationSwept(t, ctx, pool, id, ccAuditWs)
+}
--- a/server/internal/integrations/lark/channel_store.go
+++ b/server/internal/integrations/lark/channel_store.go
@@ -142,61 +142,39 @@ func (s *ChannelStore) UpsertLarkInstallation(ctx context.Context, arg UpsertIns
 	return installationFromRow(row)
 }

-// RemoveRevokedInstallationByAppID clears a revoked installation that belongs to
-// a DIFFERENT agent in the same workspace and holds the (channel_type,
-// config->>'app_id') unique slot, so the caller can re-install the same
-// Lark/Feishu app against a new agent without tripping the functional unique
-// index. The guarded DELETE is the atomic gate: it removes the row only when it
-// is still (this workspace, another agent, revoked) at delete time and RETURNS
-// its id. Only when a row is actually claimed do we clean its dependents.
-//
-//   - the SAME agent's revoked row (agent_id = agentID) never matches, so the
-//     upsert reactivates it in place (installation_id + all bindings preserved);
-//   - an ACTIVE row (bot still connected) never matches, so the install surfaces
-//     as a conflict instead of silently stealing it;
-//   - a row in ANOTHER workspace never matches.
-//
-// Keying cleanup off the RETURNING id — rather than a prior read — closes the
-// read-then-delete race: a concurrent same-agent reconnect that flips the row
-// to 'active' between a read and the delete would previously still trigger the
-// dependent cleanup on a stale id and wipe a since-reactivated installation's
-// bindings. Under READ COMMITTED the DELETE re-checks status='revoked' against
-// the live row, so it claims nothing in that case and we touch no dependents.
-//
-// channel_* has no FK/cascade (MUL-3515 §4), so every application-owned row that
-// referenced the deleted installation is cleared explicitly — chat-session
-// bindings, pending binding tokens, member links — and inbound-audit rows are
-// detached by NULLing installation_id (the app-layer stand-in for the old
-// ON DELETE SET NULL). With no FK the order is free, so cleanup runs after the
-// claiming delete; all of it is on the caller's transaction-bound queries, so
-// delete + cleanup + the follow-up upsert commit (or roll back) atomically.
-func (s *ChannelStore) RemoveRevokedInstallationByAppID(ctx context.Context, workspaceID, agentID pgtype.UUID, appID string) error {
-	installID, err := s.Queries.DeleteRevokedChannelInstallationByAppID(ctx, db.DeleteRevokedChannelInstallationByAppIDParams{
+// ReclaimDeadInstallationByAppID frees the (feishu, config->>'app_id') routing
+// slot before a rebind by removing a DEAD prior owner of the same Lark/Feishu
+// app — a revoked placeholder left by a DIFFERENT agent in this workspace, or an
+// ORPHAN whose owning workspace/agent has been deleted (#4810) — together with
+// every dependent row of that installation, in a single statement. A live owner
+// is deliberately left in place: the SAME agent's own revoked row (reactivated by
+// the follow-up upsert), and any ACTIVE owner whose agent still exists — including
+// an ARCHIVED agent, since archiving is reversible — so the upsert surfaces a
+// conflict instead of silently stealing the bot. See the full contract, and the
+// TOCTOU / EvalPlanQual reasoning, on ReclaimDeadChannelInstallationByAppID.
+func (s *ChannelStore) ReclaimDeadInstallationByAppID(ctx context.Context, workspaceID, agentID pgtype.UUID, appID string) error {
+	_, err := s.Queries.ReclaimDeadChannelInstallationByAppID(ctx, db.ReclaimDeadChannelInstallationByAppIDParams{
 		ChannelType: channelTypeFeishu,
 		AppID:       appID,
 		WorkspaceID: workspaceID,
 		AgentID:     agentID,
 	})
-	if err != nil {
-		if errors.Is(err, pgx.ErrNoRows) {
-			return nil // no blocking row claimed (same agent / active / other workspace / none)
-		}
+	if err != nil && !errors.Is(err, pgx.ErrNoRows) {
+		// pgx.ErrNoRows just means nothing was dead — a no-op, not a failure.
 		return err
 	}
+	return nil
+}

-	if err := s.Queries.DeleteChannelChatSessionBindingsByInstallation(ctx, db.DeleteChannelChatSessionBindingsByInstallationParams{
-		InstallationID: installID,
-		ChannelType:    channelTypeFeishu,
-	}); err != nil {
-		return err
-	}
-	if err := s.Queries.DeleteChannelBindingTokensByInstallation(ctx, installID); err != nil {
-		return err
-	}
-	if err := s.Queries.DeleteChannelUserBindingsByInstallation(ctx, installID); err != nil {
-		return err
-	}
-	return s.Queries.NullChannelInboundAuditInstallationID(ctx, installID)
+// InstallationOwnerByAppID returns the current owner of the (feishu, app_id)
+// routing slot so the caller can build an accurate rebind-conflict message.
+// Called after ReclaimDeadInstallationByAppID, so a row here is a live owner;
+// pgx.ErrNoRows means the slot is free.
+func (s *ChannelStore) InstallationOwnerByAppID(ctx context.Context, appID string) (db.GetChannelInstallationOwnerByAppIDRow, error) {
+	return s.Queries.GetChannelInstallationOwnerByAppID(ctx, db.GetChannelInstallationOwnerByAppIDParams{
+		ChannelType: channelTypeFeishu,
+		AppID:       appID,
+	})
 }

 func (s *ChannelStore) SetLarkInstallationStatus(ctx context.Context, arg SetInstallationStatusParams) error {
--- a/server/internal/integrations/lark/channel_store_rebind_test.go
+++ b/server/internal/integrations/lark/channel_store_rebind_test.go
@@ -8,6 +8,7 @@ import (

 	"github.com/jackc/pgx/v5"
 	"github.com/jackc/pgx/v5/pgtype"
+	"github.com/jackc/pgx/v5/pgxpool"

 	"github.com/multica-ai/multica/server/internal/util"
 	db "github.com/multica-ai/multica/server/pkg/db/generated"
@@ -18,27 +19,74 @@ import (
 // so these rows need no parent records; the test cleans up by deterministic key
 // before and after (a killed prior run must not leave colliding rows behind).
 const (
-	rbWS        = "5c09e100-0000-4000-8000-000000000001"
-	rbWS2       = "5c09e100-0000-4000-8000-000000000002"
-	rbAgentA    = "5c09e100-0000-4000-8000-00000000000a"
-	rbAgentB    = "5c09e100-0000-4000-8000-00000000000b"
-	rbInstaller = "5c09e100-0000-4000-8000-000000000005"
-	rbUser      = "5c09e100-0000-4000-8000-000000000006"
-	rbChatSess  = "5c09e100-0000-4000-8000-000000000007"
+	rbWS         = "5c09e100-0000-4000-8000-000000000001"
+	rbWS2        = "5c09e100-0000-4000-8000-000000000002"
+	rbRuntime    = "5c09e100-0000-4000-8000-000000000003"
+	rbAgentA     = "5c09e100-0000-4000-8000-00000000000a"
+	rbAgentB     = "5c09e100-0000-4000-8000-00000000000b"
+	rbAgentArch  = "5c09e100-0000-4000-8000-00000000000c"
+	rbInstaller  = "5c09e100-0000-4000-8000-000000000005"
+	rbUser       = "5c09e100-0000-4000-8000-000000000006"
+	rbChatSess   = "5c09e100-0000-4000-8000-000000000007"
+	rbGhostWS    = "5c09e100-0000-4000-8000-0000000000f1" // never seeded: a deleted workspace
+	rbGhostAgent = "5c09e100-0000-4000-8000-0000000000f2" // never seeded: a hard-deleted agent

-	rbAppSame       = "cli_rb_same"
-	rbAppDiff       = "cli_rb_diff"
-	rbAppActive     = "cli_rb_active"
-	rbAppWsFence    = "cli_rb_wsfence"
-	rbAppReactivate = "cli_rb_reactivate"
-	rbAppMove       = "cli_rb_move"
+	rbAppSame        = "cli_rb_same"
+	rbAppDiff        = "cli_rb_diff"
+	rbAppActive      = "cli_rb_active"
+	rbAppWsFence     = "cli_rb_wsfence"
+	rbAppReactivate  = "cli_rb_reactivate"
+	rbAppMove        = "cli_rb_move"
+	rbAppOrphanWS    = "cli_rb_orphan_ws"
+	rbAppOrphanAgent = "cli_rb_orphan_agent"
+	rbAppArchived    = "cli_rb_archived"
+	rbAppLive        = "cli_rb_live"
 )

-// TestChannelStore_RemoveRevokedInstallationByAppID guards the WHERE clause of
-// the DeleteRevokedChannelInstallationByAppID gate: it must claim ONLY a revoked
-// row that belongs to a DIFFERENT agent in the SAME workspace. The same agent's
-// own revoked row, any active row, and rows in another workspace must survive.
-func TestChannelStore_RemoveRevokedInstallationByAppID(t *testing.T) {
+// seedRebindOwners inserts the workspace/runtime/agent rows the rebind fixtures
+// reference. ReclaimDeadChannelInstallationByAppID now treats an installation
+// whose owning workspace OR agent row is gone as a DEAD orphan to reclaim, so
+// these tests must give their rows real owners — otherwise the orphan branch,
+// not the revoked/same-agent/other-workspace fences, would decide every case.
+// rbAgentArch is archived (a live-but-reversible owner). Idempotent; the matching
+// teardown drops the workspaces, which cascades to the runtime and agents.
+func seedRebindOwners(t *testing.T, ctx context.Context, pool *pgxpool.Pool) {
+	t.Helper()
+	exec := func(q string, args ...any) {
+		if _, err := pool.Exec(ctx, q, args...); err != nil {
+			t.Fatalf("seed rebind owner: %v", err)
+		}
+	}
+	for _, ws := range []string{rbWS, rbWS2} {
+		exec(`INSERT INTO workspace (id, name, slug, description) VALUES ($1, $2, $3, '') ON CONFLICT (id) DO NOTHING`,
+			ws, "rebind "+ws, "rebind-"+ws)
+	}
+	exec(`INSERT INTO agent_runtime (id, workspace_id, name, runtime_mode, provider)
+VALUES ($1, $2, 'rebind runtime', 'local', 'multica_daemon') ON CONFLICT (id) DO NOTHING`, rbRuntime, rbWS)
+	// Names must be unique per workspace (agent_workspace_name_unique), so key
+	// each on its id.
+	for _, agent := range []string{rbAgentA, rbAgentB} {
+		exec(`INSERT INTO agent (id, workspace_id, name, runtime_mode, runtime_id)
+VALUES ($1, $2, $3, 'local', $4) ON CONFLICT (id) DO NOTHING`, agent, rbWS, "rebind agent "+agent, rbRuntime)
+	}
+	exec(`INSERT INTO agent (id, workspace_id, name, runtime_mode, runtime_id, archived_at)
+VALUES ($1, $2, $3, 'local', $4, now()) ON CONFLICT (id) DO NOTHING`, rbAgentArch, rbWS, "rebind archived agent "+rbAgentArch, rbRuntime)
+}
+
+// cleanRebindOwners drops the seeded workspaces; the FK ON DELETE CASCADE takes
+// the runtime and agents with them. channel_installation has no such FK — the
+// bug under test — so those rows are cleaned by app_id separately.
+func cleanRebindOwners(ctx context.Context, pool *pgxpool.Pool) {
+	_, _ = pool.Exec(ctx, `DELETE FROM workspace WHERE id = ANY($1)`, []string{rbWS, rbWS2})
+}
+
+// TestChannelStore_ReclaimDeadRevokedFences guards the REVOKED branch of the
+// ReclaimDeadChannelInstallationByAppID gate: with live owners seeded, it must
+// claim ONLY a revoked row that belongs to a DIFFERENT agent in the SAME
+// workspace. The same agent's own revoked row, any active row, and a revoked row
+// in another (still-live) workspace must survive. (The orphan branch is covered
+// by TestChannelStore_ReclaimDeadReclaimsOrphansRefusesLiveOwners.)
+func TestChannelStore_ReclaimDeadRevokedFences(t *testing.T) {
 	pool := channelScopeTestDB(t)
 	ctx := context.Background()
 	store := NewChannelStore(db.New(pool))
@@ -50,6 +98,8 @@ func TestChannelStore_RemoveRevokedInstallationByAppID(t *testing.T) {
 		_, _ = pool.Exec(ctx, `DELETE FROM channel_chat_session_binding WHERE chat_session_id = $1`, rbChatSess)
 	}
 	clean()
+	seedRebindOwners(t, ctx, pool)
+	t.Cleanup(func() { cleanRebindOwners(ctx, pool) })
 	t.Cleanup(clean)

 	// insert an installation and return its id.
@@ -83,8 +133,8 @@ RETURNING id
 	t.Run("same agent revoked row is preserved", func(t *testing.T) {
 		clean()
 		id := insert(rbAppSame, rbWS, rbAgentA, "revoked")
-		if err := store.RemoveRevokedInstallationByAppID(ctx, wsUUID, agentAUUID, rbAppSame); err != nil {
-			t.Fatalf("RemoveRevokedInstallationByAppID: %v", err)
+		if err := store.ReclaimDeadInstallationByAppID(ctx, wsUUID, agentAUUID, rbAppSame); err != nil {
+			t.Fatalf("ReclaimDeadInstallationByAppID: %v", err)
 		}
 		if !exists(id) {
 			t.Fatal("same agent's own revoked row was deleted; it must be reactivated in place by the upsert, not orphaned")
@@ -94,8 +144,8 @@ RETURNING id
 	t.Run("different agent revoked row is deleted", func(t *testing.T) {
 		clean()
 		id := insert(rbAppDiff, rbWS, rbAgentA, "revoked")
-		if err := store.RemoveRevokedInstallationByAppID(ctx, wsUUID, agentBUUID, rbAppDiff); err != nil {
-			t.Fatalf("RemoveRevokedInstallationByAppID: %v", err)
+		if err := store.ReclaimDeadInstallationByAppID(ctx, wsUUID, agentBUUID, rbAppDiff); err != nil {
+			t.Fatalf("ReclaimDeadInstallationByAppID: %v", err)
 		}
 		if exists(id) {
 			t.Fatal("a different agent's revoked row was not deleted; it would keep blocking the app_id unique slot")
@@ -105,8 +155,8 @@ RETURNING id
 	t.Run("active row is never deleted", func(t *testing.T) {
 		clean()
 		id := insert(rbAppActive, rbWS, rbAgentA, "active")
-		if err := store.RemoveRevokedInstallationByAppID(ctx, wsUUID, agentBUUID, rbAppActive); err != nil {
-			t.Fatalf("RemoveRevokedInstallationByAppID: %v", err)
+		if err := store.ReclaimDeadInstallationByAppID(ctx, wsUUID, agentBUUID, rbAppActive); err != nil {
+			t.Fatalf("ReclaimDeadInstallationByAppID: %v", err)
 		}
 		if !exists(id) {
 			t.Fatal("an active installation was deleted through the revoked-cleanup path")
@@ -116,8 +166,8 @@ RETURNING id
 	t.Run("other workspace revoked row is preserved", func(t *testing.T) {
 		clean()
 		id := insert(rbAppWsFence, rbWS2, rbAgentA, "revoked")
-		if err := store.RemoveRevokedInstallationByAppID(ctx, wsUUID, agentBUUID, rbAppWsFence); err != nil {
-			t.Fatalf("RemoveRevokedInstallationByAppID: %v", err)
+		if err := store.ReclaimDeadInstallationByAppID(ctx, wsUUID, agentBUUID, rbAppWsFence); err != nil {
+			t.Fatalf("ReclaimDeadInstallationByAppID: %v", err)
 		}
 		if !exists(id) {
 			t.Fatal("a revoked row in another workspace was deleted; the delete must stay workspace-scoped")
@@ -145,6 +195,8 @@ func TestChannelStore_ReinstallReactivationSemantics(t *testing.T) {
 		_, _ = pool.Exec(ctx, `DELETE FROM channel_chat_session_binding WHERE chat_session_id = $1`, rbChatSess)
 	}
 	clean()
+	seedRebindOwners(t, ctx, pool)
+	t.Cleanup(func() { cleanRebindOwners(ctx, pool) })
 	t.Cleanup(clean)

 	insertRevoked := func(app, agent string) pgtype.UUID {
@@ -207,7 +259,7 @@ VALUES ($1, $2, 'feishu', 'oc_rb_chat', 'p2p')

 		// finishSuccess order: cleanup for the current agent (a no-op for the
 		// same agent), then upsert.
-		if err := store.RemoveRevokedInstallationByAppID(ctx, util.MustParseUUID(rbWS), util.MustParseUUID(rbAgentA), rbAppReactivate); err != nil {
+		if err := store.ReclaimDeadInstallationByAppID(ctx, util.MustParseUUID(rbWS), util.MustParseUUID(rbAgentA), rbAppReactivate); err != nil {
 			t.Fatalf("cleanup: %v", err)
 		}
 		inst := upsert(rbAgentA, rbAppReactivate)
@@ -228,7 +280,7 @@ VALUES ($1, $2, 'feishu', 'oc_rb_chat', 'p2p')
 		oldID := insertRevoked(rbAppMove, rbAgentA)
 		attachBindings(oldID)

-		if err := store.RemoveRevokedInstallationByAppID(ctx, util.MustParseUUID(rbWS), util.MustParseUUID(rbAgentB), rbAppMove); err != nil {
+		if err := store.ReclaimDeadInstallationByAppID(ctx, util.MustParseUUID(rbWS), util.MustParseUUID(rbAgentB), rbAppMove); err != nil {
 			t.Fatalf("cleanup: %v", err)
 		}
 		inst := upsert(rbAgentB, rbAppMove)
@@ -269,6 +321,8 @@ func TestChannelStore_RebindCleansDependentRows(t *testing.T) {
 		_, _ = pool.Exec(ctx, `DELETE FROM channel_inbound_audit WHERE channel_event_id = $1`, auditEvent)
 	}
 	clean()
+	seedRebindOwners(t, ctx, pool)
+	t.Cleanup(func() { cleanRebindOwners(ctx, pool) })
 	t.Cleanup(clean)

 	// A revoked installation for agent A carrying the full spread of dependents.
@@ -295,8 +349,8 @@ VALUES ($1, $2, $3, 'feishu', 'ou_rb_user', now() + interval '10 minutes')`, tok
 VALUES ($1, 'feishu', 'im.message.receive_v1', $2, 'revoked_installation')`, oldID, auditEvent)

 	// Rebind the app to a DIFFERENT agent.
-	if err := store.RemoveRevokedInstallationByAppID(ctx, util.MustParseUUID(rbWS), util.MustParseUUID(rbAgentB), app); err != nil {
-		t.Fatalf("RemoveRevokedInstallationByAppID: %v", err)
+	if err := store.ReclaimDeadInstallationByAppID(ctx, util.MustParseUUID(rbWS), util.MustParseUUID(rbAgentB), app); err != nil {
+		t.Fatalf("ReclaimDeadInstallationByAppID: %v", err)
 	}

 	count := func(q string, args ...any) int {
@@ -335,7 +389,7 @@ VALUES ($1, 'feishu', 'im.message.receive_v1', $2, 'revoked_installation')`, old
 //   - txReconnect (agent A reconnecting to the SAME agent) reactivates the row
 //     to 'active' but holds the row lock uncommitted;
 //   - txRebind (agent B rebinding to a DIFFERENT agent) runs the full cleanup
-//     via RemoveRevokedInstallationByAppID.
+//     via ReclaimDeadInstallationByAppID.
 //
 // The old read-then-clean-then-delete shape would read the still-committed
 // 'revoked' row, wipe its dependents, then no-op on the fenced delete — losing
@@ -363,6 +417,8 @@ func TestChannelStore_RebindGuardedDeleteRaceWithReactivation(t *testing.T) {
 		_, _ = pool.Exec(ctx, `DELETE FROM channel_inbound_audit WHERE channel_event_id = $1`, auditEvent)
 	}
 	clean()
+	seedRebindOwners(t, ctx, pool)
+	t.Cleanup(func() { cleanRebindOwners(ctx, pool) })
 	t.Cleanup(clean)

 	// A revoked installation for agent A, with the full spread of dependents.
@@ -408,7 +464,7 @@ VALUES ($1, 'feishu', 'im.message.receive_v1', $2, 'revoked_installation')`, idS
 			return
 		}
 		defer txRebind.Rollback(ctx)
-		if err := store.WithTx(txRebind).RemoveRevokedInstallationByAppID(ctx, util.MustParseUUID(rbWS), util.MustParseUUID(rbAgentB), app); err != nil {
+		if err := store.WithTx(txRebind).ReclaimDeadInstallationByAppID(ctx, util.MustParseUUID(rbWS), util.MustParseUUID(rbAgentB), app); err != nil {
 			done <- err
 			return
 		}
@@ -452,3 +508,110 @@ VALUES ($1, 'feishu', 'im.message.receive_v1', $2, 'revoked_installation')`, idS
 		t.Fatalf("audit reference detached by the racing rebind: got %d, want 1", n)
 	}
 }
+
+// TestChannelStore_ReclaimDeadReclaimsOrphansRefusesLiveOwners covers the ORPHAN
+// branch of the reclaim gate — the #4810 fix. An installation whose owning
+// workspace or agent has been hard-deleted is a dead orphan still occupying the
+// (channel_type, app_id) routing slot; the reclaim must clear it so the bot can
+// be rebound. A live owner — an active agent, INCLUDING an archived one — must be
+// left in place so the follow-up upsert refuses the rebind rather than stealing
+// the bot, and InstallationOwnerByAppID must report who holds it (so the caller
+// can build an accurate conflict message).
+func TestChannelStore_ReclaimDeadReclaimsOrphansRefusesLiveOwners(t *testing.T) {
+	pool := channelScopeTestDB(t)
+	ctx := context.Background()
+	store := NewChannelStore(db.New(pool))
+
+	apps := []string{rbAppOrphanWS, rbAppOrphanAgent, rbAppLive, rbAppArchived}
+	clean := func() {
+		_, _ = pool.Exec(ctx, `DELETE FROM channel_installation WHERE config->>'app_id' = ANY($1)`, apps)
+	}
+	clean()
+	seedRebindOwners(t, ctx, pool)
+	t.Cleanup(func() { cleanRebindOwners(ctx, pool) })
+	t.Cleanup(clean)
+
+	insert := func(app, ws, agent, status string) pgtype.UUID {
+		var id string
+		if err := pool.QueryRow(ctx, `
+INSERT INTO channel_installation (workspace_id, agent_id, channel_type, config, installer_user_id, status)
+VALUES ($1, $2, 'feishu', jsonb_build_object('app_id', $3::text), $4, $5)
+RETURNING id
+`, ws, agent, app, rbInstaller, status).Scan(&id); err != nil {
+			t.Fatalf("insert installation app=%s: %v", app, err)
+		}
+		return util.MustParseUUID(id)
+	}
+	exists := func(id pgtype.UUID) bool {
+		_, err := store.GetLarkInstallation(ctx, id)
+		if err == nil {
+			return true
+		}
+		if errors.Is(err, pgx.ErrNoRows) {
+			return false
+		}
+		t.Fatalf("GetLarkInstallation: %v", err)
+		return false
+	}
+	// A NEW agent B, in a live workspace, is the one rebinding each app.
+	wsUUID := util.MustParseUUID(rbWS)
+	agentBUUID := util.MustParseUUID(rbAgentB)
+
+	t.Run("orphan from a deleted workspace is reclaimed", func(t *testing.T) {
+		clean()
+		id := insert(rbAppOrphanWS, rbGhostWS, rbAgentA, "active")
+		if err := store.ReclaimDeadInstallationByAppID(ctx, wsUUID, agentBUUID, rbAppOrphanWS); err != nil {
+			t.Fatalf("ReclaimDeadInstallationByAppID: %v", err)
+		}
+		if exists(id) {
+			t.Fatal("an installation whose workspace no longer exists was not reclaimed; the bot would stay un-rebindable")
+		}
+	})
+
+	t.Run("orphan from a hard-deleted agent is reclaimed", func(t *testing.T) {
+		clean()
+		id := insert(rbAppOrphanAgent, rbWS, rbGhostAgent, "active")
+		if err := store.ReclaimDeadInstallationByAppID(ctx, wsUUID, agentBUUID, rbAppOrphanAgent); err != nil {
+			t.Fatalf("ReclaimDeadInstallationByAppID: %v", err)
+		}
+		if exists(id) {
+			t.Fatal("an installation whose agent was hard-deleted was not reclaimed")
+		}
+	})
+
+	t.Run("live active owner is refused, not stolen", func(t *testing.T) {
+		clean()
+		id := insert(rbAppLive, rbWS, rbAgentA, "active")
+		if err := store.ReclaimDeadInstallationByAppID(ctx, wsUUID, agentBUUID, rbAppLive); err != nil {
+			t.Fatalf("ReclaimDeadInstallationByAppID: %v", err)
+		}
+		if !exists(id) {
+			t.Fatal("a live active owner was reclaimed; agent B would silently steal the bot")
+		}
+		owner, err := store.InstallationOwnerByAppID(ctx, rbAppLive)
+		if err != nil {
+			t.Fatalf("InstallationOwnerByAppID: %v", err)
+		}
+		if owner.WorkspaceID != wsUUID || owner.AgentID != util.MustParseUUID(rbAgentA) || owner.AgentArchivedAt.Valid {
+			t.Fatalf("owner mismatch: ws=%v agent=%v archived=%v", owner.WorkspaceID, owner.AgentID, owner.AgentArchivedAt.Valid)
+		}
+	})
+
+	t.Run("archived agent owner is refused and reported archived", func(t *testing.T) {
+		clean()
+		id := insert(rbAppArchived, rbWS, rbAgentArch, "active")
+		if err := store.ReclaimDeadInstallationByAppID(ctx, wsUUID, agentBUUID, rbAppArchived); err != nil {
+			t.Fatalf("ReclaimDeadInstallationByAppID: %v", err)
+		}
+		if !exists(id) {
+			t.Fatal("an archived agent's installation was reclaimed; archiving is reversible so the bot must stay owned")
+		}
+		owner, err := store.InstallationOwnerByAppID(ctx, rbAppArchived)
+		if err != nil {
+			t.Fatalf("InstallationOwnerByAppID: %v", err)
+		}
+		if !owner.AgentArchivedAt.Valid {
+			t.Fatal("owner lookup did not report the agent as archived; the conflict message could not distinguish it")
+		}
+	})
+}
--- a/server/internal/integrations/lark/registration_service.go
+++ b/server/internal/integrations/lark/registration_service.go
@@ -11,12 +11,18 @@ import (
 	"sync"
 	"time"

+	"github.com/jackc/pgx/v5/pgconn"
 	"github.com/jackc/pgx/v5/pgtype"
 	"github.com/multica-ai/multica/server/internal/events"
 	db "github.com/multica-ai/multica/server/pkg/db/generated"
 	"github.com/multica-ai/multica/server/pkg/protocol"
 )

+// pgUniqueViolation is the Postgres SQLSTATE for a unique-constraint violation.
+// A rebind upsert that trips the (channel_type, config->>'app_id') index after
+// the dead-owner reclaim ran means a LIVE owner still holds the slot.
+const pgUniqueViolation = "23505"
+
 // RegistrationSessionStatus is the discriminated state a `begin`
 // session lives in. The HTTP status endpoint serializes the underlying
 // string verbatim so the frontend can pattern-match without parsing
@@ -571,17 +577,17 @@ func (s *RegistrationService) finishSuccess(ctx context.Context, sess *registrat
 	defer tx.Rollback(ctx)
 	qtx := s.queries.WithTx(tx)

-	// If the same Feishu app (app_id) was previously bound to a DIFFERENT
-	// agent in this workspace and later revoked, that revoked row still
-	// holds the (channel_type, config->>'app_id') unique index slot and
-	// blocks the UpsertChannelInstallation INSERT below. Remove the
-	// revoked placeholder first — the transaction wraps both the delete
-	// and the upsert so a failure between them rolls back cleanly. The
-	// current agent (sess.agentID) is excluded: re-connecting the SAME
-	// agent reactivates its own revoked row in place via the upsert's
-	// ON CONFLICT, keeping its installation_id and every binding intact.
-	if err := qtx.RemoveRevokedInstallationByAppID(ctx, sess.workspaceID, sess.agentID, res.ClientID); err != nil {
-		s.cfg.Logger.Warn("lark registration: cleanup revoked installation",
+	// If the same Feishu app (app_id) is held by a DEAD prior owner — a revoked
+	// placeholder left by a DIFFERENT agent in this workspace, or an orphan whose
+	// workspace/agent was deleted (#4810) — that row still occupies the
+	// (channel_type, config->>'app_id') unique slot and blocks the
+	// UpsertChannelInstallation INSERT below. Reclaim it first — the transaction
+	// wraps both the delete and the upsert so a failure between them rolls back
+	// cleanly. A live owner is left in place (the SAME agent's own revoked row is
+	// reactivated in place by the upsert; an active/archived agent stays owned),
+	// so the upsert surfaces the conflict below instead of stealing the bot.
+	if err := qtx.ReclaimDeadInstallationByAppID(ctx, sess.workspaceID, sess.agentID, res.ClientID); err != nil {
+		s.cfg.Logger.Warn("lark registration: reclaim dead installation",
 			"session_id", sess.id, "err", err)
 		sess.markError(RegistrationReasonInternalError, err.Error(), s.gcDeadline())
 		return
@@ -600,7 +606,16 @@ func (s *RegistrationService) finishSuccess(ctx context.Context, sess *registrat
 	if err != nil {
 		s.cfg.Logger.Warn("lark registration: upsert installation",
 			"session_id", sess.id, "err", err)
-		sess.markError(RegistrationReasonInstallationConflict, err.Error(), s.gcDeadline())
+		// A unique violation here means the app_id slot is held by a LIVE owner
+		// (the reclaim above already cleared every dead one). Surface who holds
+		// it — another agent in this workspace, an archived agent, or a different
+		// workspace — instead of leaking the raw Postgres error.
+		msg := err.Error()
+		var pgErr *pgconn.PgError
+		if errors.As(err, &pgErr) && pgErr.Code == pgUniqueViolation {
+			msg = s.liveOwnerConflictMessage(ctx, sess.workspaceID, res.ClientID)
+		}
+		sess.markError(RegistrationReasonInstallationConflict, msg, s.gcDeadline())
 		return
 	}

@@ -634,6 +649,28 @@ func (s *RegistrationService) finishSuccess(ctx context.Context, sess *registrat
 		"installation_id", uuidString(inst.ID))
 }

+// liveOwnerConflictMessage builds the user-facing copy for a rebind refused
+// because the Feishu app's routing slot is held by a LIVE owner. It names which
+// kind of owner so the user knows how to recover, instead of the old catch-all
+// "connected to a different Multica workspace" that lied when the real owner sat
+// in the SAME workspace (#4810). Looked up on the base pool, not the aborted
+// upsert tx. If the slot turns out free (a concurrent disconnect between the
+// upsert and this read), a generic message is enough — the user can just retry.
+func (s *RegistrationService) liveOwnerConflictMessage(ctx context.Context, requestingWorkspaceID pgtype.UUID, appID string) string {
+	owner, err := s.queries.InstallationOwnerByAppID(ctx, appID)
+	if err != nil {
+		return "This Feishu app is already connected to another agent. Disconnect it there first, then connect it here."
+	}
+	switch {
+	case owner.WorkspaceID != requestingWorkspaceID:
+		return "This Feishu app is already connected to a different Multica workspace. Disconnect it there before connecting it here."
+	case owner.AgentArchivedAt.Valid:
+		return "This Feishu app is connected to an archived agent in this workspace. Restore that agent, or disconnect its bot, before connecting it here."
+	default:
+		return "This Feishu app is already connected to another agent in this workspace. Disconnect it there first, then connect it here."
+	}
+}
+
 func (s *RegistrationService) gcDeadline() time.Time {
 	return s.cfg.Now().Add(s.cfg.SessionTTL)
 }
--- a/server/internal/integrations/slack/byo_install.go
+++ b/server/internal/integrations/slack/byo_install.go
@@ -123,6 +123,7 @@ func (s *InstallService) RegisterBYO(ctx context.Context, p RegisterBYOParams) (
 		wsID:        p.WorkspaceID,
 		agentID:     p.AgentID,
 		installerID: p.InitiatorID,
+		appIDKey:    appID,
 		configJSON:  cfgJSON,
 	})
 }
--- a/server/internal/integrations/slack/byo_install_test.go
+++ b/server/internal/integrations/slack/byo_install_test.go
@@ -196,15 +196,17 @@ func TestRegisterBYO_AuthTestFailure(t *testing.T) {
 	}
 }

-func TestRegisterBYO_AppAlreadyConnected_Rejected(t *testing.T) {
+func TestRegisterBYO_AppConnectedToAnotherWorkspace_Rejected(t *testing.T) {
 	srv := authTestServer(t, true)
 	defer srv.Close()
-	// The pasted app is already connected to another agent / workspace, so the
-	// (channel_type, app_id) routing index rejects the upsert (unique violation).
-	// We must refuse, not steal it.
+	// The pasted app is live-owned by an agent in a DIFFERENT Multica workspace,
+	// so after the dead-owner reclaim the (channel_type, app_id) routing index
+	// still rejects the upsert. We must refuse, not steal it — and name the real
+	// case (another workspace), not the old catch-all.
 	q := &fakeInstallQueries{
-		rowID:      mustUUID(t, "44444444-4444-4444-4444-444444444444"),
-		appIDTaken: true,
+		rowID:            mustUUID(t, "44444444-4444-4444-4444-444444444444"),
+		appIDTaken:       true,
+		ownerWorkspaceID: mustUUID(t, "99999999-9999-9999-9999-999999999999"),
 	}
 	svc := newTestInstallService(t, q)
 	svc.apiURL = srv.URL + "/"
@@ -215,6 +217,55 @@ func TestRegisterBYO_AppAlreadyConnected_Rejected(t *testing.T) {
 	)); err != ErrTeamOwnedByAnotherWorkspace {
 		t.Fatalf("app already connected = %v, want ErrTeamOwnedByAnotherWorkspace", err)
 	}
+	if !q.reclaimCalled {
+		t.Error("install must attempt a dead-owner reclaim before the upsert")
+	}
+}
+
+func TestRegisterBYO_AppConnectedToAnotherAgentSameWorkspace_Rejected(t *testing.T) {
+	srv := authTestServer(t, true)
+	defer srv.Close()
+	// The pasted app is live-owned by a DIFFERENT (non-archived) agent in the
+	// SAME workspace. The old catch-all wrongly blamed "another workspace"; this
+	// must surface the same-workspace sentinel so the UI points at the Disconnect
+	// the user can actually reach (#4810).
+	q := &fakeInstallQueries{
+		rowID:            mustUUID(t, "44444444-4444-4444-4444-444444444444"),
+		appIDTaken:       true,
+		ownerWorkspaceID: mustUUID(t, "11111111-1111-1111-1111-111111111111"),
+	}
+	svc := newTestInstallService(t, q)
+	svc.apiURL = srv.URL + "/"
+
+	if _, err := svc.RegisterBYO(context.Background(), byoParams(
+		"11111111-1111-1111-1111-111111111111",
+		"22222222-2222-2222-2222-222222222222",
+	)); err != ErrTeamOwnedBySameWorkspace {
+		t.Fatalf("app owned by another agent in this workspace = %v, want ErrTeamOwnedBySameWorkspace", err)
+	}
+}
+
+func TestRegisterBYO_AppConnectedToArchivedAgent_Rejected(t *testing.T) {
+	srv := authTestServer(t, true)
+	defer srv.Close()
+	// The pasted app's owning agent is archived — a live-but-reversible owner —
+	// so the reclaim leaves it in place and the upsert is refused. The user is
+	// told to restore the agent or disconnect its bot, not that it's gone.
+	q := &fakeInstallQueries{
+		rowID:            mustUUID(t, "44444444-4444-4444-4444-444444444444"),
+		appIDTaken:       true,
+		ownerWorkspaceID: mustUUID(t, "11111111-1111-1111-1111-111111111111"),
+		ownerArchived:    true,
+	}
+	svc := newTestInstallService(t, q)
+	svc.apiURL = srv.URL + "/"
+
+	if _, err := svc.RegisterBYO(context.Background(), byoParams(
+		"11111111-1111-1111-1111-111111111111",
+		"22222222-2222-2222-2222-222222222222",
+	)); err != ErrTeamOwnedByArchivedAgent {
+		t.Fatalf("app owned by an archived agent = %v, want ErrTeamOwnedByArchivedAgent", err)
+	}
 }

 func TestRegisterBYO_ReconnectSameAgent_UpdatesRowInPlace(t *testing.T) {
--- a/server/internal/integrations/slack/install.go
+++ b/server/internal/integrations/slack/install.go
@@ -28,11 +28,20 @@ var (
 	// ErrInstallationNotFound surfaces "no row matches in this workspace".
 	ErrInstallationNotFound = errors.New("slack installation not found")
 	// ErrTeamOwnedByAnotherWorkspace is returned when the pasted Slack app is
-	// already connected to a DIFFERENT agent or Multica workspace — it would
-	// collide with the (channel_type, app_id) routing index. A Slack app is one
-	// bot identity and maps to one agent; reusing it elsewhere requires
-	// disconnecting it there first.
-	ErrTeamOwnedByAnotherWorkspace = errors.New("slack: this Slack app is already connected to another agent or Multica workspace")
+	// already connected to a live owner in a DIFFERENT Multica workspace — it
+	// would collide with the (channel_type, app_id) routing index. A Slack app is
+	// one bot identity and maps to one agent; reusing it here requires
+	// disconnecting it in the other workspace first.
+	ErrTeamOwnedByAnotherWorkspace = errors.New("slack: this Slack app is already connected to a different Multica workspace")
+	// ErrTeamOwnedBySameWorkspace is returned when the app is already connected to
+	// a DIFFERENT (live, non-archived) agent in the SAME workspace. The old
+	// catch-all wrongly blamed "another workspace"; naming the same-workspace case
+	// points the user at the Disconnect they can actually reach (#4810).
+	ErrTeamOwnedBySameWorkspace = errors.New("slack: this Slack app is already connected to another agent in this workspace")
+	// ErrTeamOwnedByArchivedAgent is returned when the app's owning agent is
+	// archived (and so still holds the bot, since archiving is reversible). The
+	// user recovers by restoring that agent or disconnecting its bot.
+	ErrTeamOwnedByArchivedAgent = errors.New("slack: this Slack app is connected to an archived agent in this workspace")
 )

 // installQueries is the slice of generated queries InstallService needs. WithTx
@@ -41,6 +50,8 @@ var (
 type installQueries interface {
 	WithTx(tx pgx.Tx) installQueries
 	UpsertChannelInstallation(ctx context.Context, arg db.UpsertChannelInstallationParams) (db.ChannelInstallation, error)
+	ReclaimDeadChannelInstallationByAppID(ctx context.Context, arg db.ReclaimDeadChannelInstallationByAppIDParams) (pgtype.UUID, error)
+	GetChannelInstallationOwnerByAppID(ctx context.Context, arg db.GetChannelInstallationOwnerByAppIDParams) (db.GetChannelInstallationOwnerByAppIDRow, error)
 	ListChannelInstallationsByWorkspace(ctx context.Context, arg db.ListChannelInstallationsByWorkspaceParams) ([]db.ChannelInstallation, error)
 	GetChannelInstallationInWorkspace(ctx context.Context, arg db.GetChannelInstallationInWorkspaceParams) (db.ChannelInstallation, error)
 	SetChannelInstallationStatus(ctx context.Context, arg db.SetChannelInstallationStatusParams) error
@@ -115,6 +126,10 @@ type installPersist struct {
 	wsID        pgtype.UUID
 	agentID     pgtype.UUID
 	installerID pgtype.UUID
+	// appIDKey is the Slack app id stored at config->>'app_id'; it MUST equal the
+	// app_id inside configJSON. It keys the dead-owner reclaim and the live-owner
+	// lookup that drives the accurate conflict message.
+	appIDKey string
 	// configJSON holds the Slack app id (config->>'app_id') used for inbound
 	// routing; the ROW itself is keyed by (workspace, agent) — one bot per agent.
 	configJSON []byte
@@ -142,6 +157,21 @@ func (s *InstallService) persistInstall(ctx context.Context, p installPersist) (
 	defer func() { _ = tx.Rollback(ctx) }()
 	qtx := s.q.WithTx(tx)

+	// Free the (slack, app_id) routing slot from any DEAD prior owner — a revoked
+	// placeholder, or an orphan whose owning workspace/agent was deleted (#4810) —
+	// before the upsert, so a bot whose old owner is gone can be rebound. A live
+	// owner (active agent, including an archived one) is left in place and trips
+	// the unique index below, which we turn into an accurate conflict.
+	if _, err := qtx.ReclaimDeadChannelInstallationByAppID(ctx, db.ReclaimDeadChannelInstallationByAppIDParams{
+		ChannelType: string(TypeSlack),
+		AppID:       p.appIDKey,
+		WorkspaceID: p.wsID,
+		AgentID:     p.agentID,
+	}); err != nil && !errors.Is(err, pgx.ErrNoRows) {
+		// pgx.ErrNoRows just means nothing was dead — a no-op, not a failure.
+		return db.ChannelInstallation{}, fmt.Errorf("reclaim dead slack installation: %w", err)
+	}
+
 	inst, err := qtx.UpsertChannelInstallation(ctx, db.UpsertChannelInstallationParams{
 		WorkspaceID:     p.wsID,
 		AgentID:         p.agentID,
@@ -152,7 +182,7 @@ func (s *InstallService) persistInstall(ctx context.Context, p installPersist) (
 	if err != nil {
 		var pgErr *pgconn.PgError
 		if errors.As(err, &pgErr) && pgErr.Code == pgUniqueViolation {
-			return db.ChannelInstallation{}, ErrTeamOwnedByAnotherWorkspace
+			return db.ChannelInstallation{}, s.liveOwnerConflictErr(ctx, p.wsID, p.appIDKey)
 		}
 		return db.ChannelInstallation{}, fmt.Errorf("upsert slack installation: %w", err)
 	}
@@ -162,6 +192,31 @@ func (s *InstallService) persistInstall(ctx context.Context, p installPersist) (
 	return inst, nil
 }

+// liveOwnerConflictErr classifies who holds the (slack, app_id) routing slot
+// after the dead-owner reclaim ran, so persistInstall returns a sentinel the
+// handler renders as an accurate message rather than the old catch-all that
+// always blamed "another workspace" (#4810). Read on the base pool (s.q), since
+// the failed upsert has aborted the tx. A now-free slot (concurrent disconnect)
+// or lookup error falls back to the generic cross-workspace sentinel — a retry
+// then succeeds.
+func (s *InstallService) liveOwnerConflictErr(ctx context.Context, requestingWorkspaceID pgtype.UUID, appID string) error {
+	owner, err := s.q.GetChannelInstallationOwnerByAppID(ctx, db.GetChannelInstallationOwnerByAppIDParams{
+		ChannelType: string(TypeSlack),
+		AppID:       appID,
+	})
+	if err != nil {
+		return ErrTeamOwnedByAnotherWorkspace
+	}
+	switch {
+	case owner.WorkspaceID != requestingWorkspaceID:
+		return ErrTeamOwnedByAnotherWorkspace
+	case owner.AgentArchivedAt.Valid:
+		return ErrTeamOwnedByArchivedAgent
+	default:
+		return ErrTeamOwnedBySameWorkspace
+	}
+}
+
 // ListByWorkspace returns every Slack installation in the workspace (active and
 // revoked), for the management surface.
 func (s *InstallService) ListByWorkspace(ctx context.Context, wsID pgtype.UUID) ([]db.ChannelInstallation, error) {
--- a/server/internal/integrations/slack/install_test.go
+++ b/server/internal/integrations/slack/install_test.go
@@ -41,16 +41,46 @@ type fakeInstallQueries struct {
 	existing *db.ChannelInstallation
 	// appIDTaken makes UpsertChannelInstallation report a unique-constraint
 	// violation on the (channel_type, app_id) routing index — i.e. the pasted app
-	// is already connected to another agent / workspace.
+	// is already connected to a LIVE owner (the reclaim has run by then).
 	appIDTaken   bool
 	upsertParams db.UpsertChannelInstallationParams
 	upsertCalled bool
 	rowID        pgtype.UUID
+
+	// reclaimedID, when set, is returned by ReclaimDeadChannelInstallationByAppID
+	// to model a dead prior owner having been cleared; otherwise it reports
+	// pgx.ErrNoRows (nothing was dead). reclaimCalled records that the install
+	// path ran the reclaim before upserting.
+	reclaimedID   *pgtype.UUID
+	reclaimCalled bool
+	// ownerWorkspaceID / ownerArchived / ownerMissing drive the live-owner lookup
+	// that classifies an appIDTaken conflict into the right sentinel.
+	ownerWorkspaceID pgtype.UUID
+	ownerArchived    bool
+	ownerMissing     bool
 }

 // WithTx returns the same fake — the fake tx is a no-op token.
 func (f *fakeInstallQueries) WithTx(_ pgx.Tx) installQueries { return f }

+func (f *fakeInstallQueries) ReclaimDeadChannelInstallationByAppID(_ context.Context, _ db.ReclaimDeadChannelInstallationByAppIDParams) (pgtype.UUID, error) {
+	f.reclaimCalled = true
+	if f.reclaimedID != nil {
+		return *f.reclaimedID, nil
+	}
+	return pgtype.UUID{}, pgx.ErrNoRows
+}
+
+func (f *fakeInstallQueries) GetChannelInstallationOwnerByAppID(_ context.Context, _ db.GetChannelInstallationOwnerByAppIDParams) (db.GetChannelInstallationOwnerByAppIDRow, error) {
+	if f.ownerMissing {
+		return db.GetChannelInstallationOwnerByAppIDRow{}, pgx.ErrNoRows
+	}
+	return db.GetChannelInstallationOwnerByAppIDRow{
+		WorkspaceID:     f.ownerWorkspaceID,
+		AgentArchivedAt: pgtype.Timestamptz{Valid: f.ownerArchived},
+	}, nil
+}
+
 func (f *fakeInstallQueries) UpsertChannelInstallation(_ context.Context, arg db.UpsertChannelInstallationParams) (db.ChannelInstallation, error) {
 	f.upsertCalled = true
 	f.upsertParams = arg
--- a/server/pkg/db/generated/channel.sql.go
+++ b/server/pkg/db/generated/channel.sql.go
@@ -400,6 +400,41 @@ func (q *Queries) DeleteChannelChatSessionBindingsByInstallation(ctx context.Con
 	return err
 }

+const deleteChannelInstallationsByArchivedRuntimeAgents = `-- name: DeleteChannelInstallationsByArchivedRuntimeAgents :exec
+WITH doomed AS (
+    SELECT id FROM channel_installation
+    WHERE agent_id IN (
+        SELECT id FROM agent WHERE runtime_id = $1 AND archived_at IS NOT NULL
+    )
+),
+cleared_chat_sessions AS (
+    DELETE FROM channel_chat_session_binding WHERE installation_id IN (SELECT id FROM doomed)
+),
+cleared_binding_tokens AS (
+    DELETE FROM channel_binding_token WHERE installation_id IN (SELECT id FROM doomed)
+),
+cleared_user_bindings AS (
+    DELETE FROM channel_user_binding WHERE installation_id IN (SELECT id FROM doomed)
+),
+detached_audit AS (
+    UPDATE channel_inbound_audit SET installation_id = NULL WHERE installation_id IN (SELECT id FROM doomed)
+)
+DELETE FROM channel_installation WHERE id IN (SELECT id FROM doomed)
+`
+
+// Application-layer replacement for the (deliberately absent, MUL-3515 §4)
+// workspace/agent ON DELETE CASCADE: on runtime teardown, before the archived
+// agents are hard-deleted, remove every channel installation they own — plus all
+// of each installation's dependent rows — so no orphaned installation keeps
+// occupying its bot's (channel_type, app_id) routing slot after its agent is gone
+// (#4810). MUST run in the same tx as, and BEFORE, DeleteArchivedAgentsByRuntime.
+// Mirrors the agent hard-delete predicate (runtime_id, archived_at IS NOT NULL)
+// exactly.
+func (q *Queries) DeleteChannelInstallationsByArchivedRuntimeAgents(ctx context.Context, runtimeID pgtype.UUID) error {
+	_, err := q.db.Exec(ctx, deleteChannelInstallationsByArchivedRuntimeAgents, runtimeID)
+	return err
+}
+
 const deleteChannelUserBindingsByInstallation = `-- name: DeleteChannelUserBindingsByInstallation :exec
 DELETE FROM channel_user_binding
 WHERE installation_id = $1
@@ -436,61 +471,6 @@ func (q *Queries) DeleteChannelUserBindingsByWorkspaceMember(ctx context.Context
 	return err
 }

-const deleteRevokedChannelInstallationByAppID = `-- name: DeleteRevokedChannelInstallationByAppID :one
-DELETE FROM channel_installation
-WHERE channel_type = $1
-  AND config ->> 'app_id' = $2::text
-  AND workspace_id = $3
-  AND agent_id <> $4
-  AND status = 'revoked'
-RETURNING id
-`
-
-type DeleteRevokedChannelInstallationByAppIDParams struct {
-	ChannelType string      `json:"channel_type"`
-	AppID       string      `json:"app_id"`
-	WorkspaceID pgtype.UUID `json:"workspace_id"`
-	AgentID     pgtype.UUID `json:"agent_id"`
-}
-
-// Hard-delete a REVOKED installation that belongs to a DIFFERENT agent, keyed
-// by its platform app identity, and RETURN the deleted id. When a Feishu/Lark
-// Bot is disconnected from agent A (status → 'revoked') and the same Bot (same
-// app_id) is later bound to a DIFFERENT agent B, agent A's revoked row still
-// occupies the (channel_type, config->>'app_id') unique slot and blocks the
-// UpsertChannelInstallation INSERT for B. Removing that placeholder first lets
-// the upsert create a fresh row for the new agent.
-//
-// The `agent_id <> arg` fence is load-bearing: re-connecting the SAME agent
-// must NOT delete its own revoked row. UpsertChannelInstallation conflicts on
-// (workspace_id, agent_id, channel_type), so the same agent's revoked row is
-// reactivated in place (status → 'active') with its installation_id — and every
-// channel_user_binding / channel_chat_session_binding hanging off it —
-// preserved. Deleting it here would force an INSERT with a fresh installation_id
-// and orphan all of that agent's member links and chat sessions.
-//
-// Fenced to one workspace AND status='revoked' so an active installation can
-// never be silently deleted through this path.
-//
-// This DELETE is the atomic gate for the whole rebind cleanup: the caller keys
-// its dependent-row cleanup off the RETURNING id, so cleanup runs ONLY for a row
-// this statement actually removed. Under READ COMMITTED, a concurrent same-agent
-// reconnect that reactivates the row to 'active' first makes the WHERE re-check
-// fail (EvalPlanQual), this deletes nothing (pgx.ErrNoRows), and no dependents
-// are touched — closing the read-then-delete TOCTOU where a stale "it was
-// revoked" read could wipe the bindings of a since-reactivated installation.
-func (q *Queries) DeleteRevokedChannelInstallationByAppID(ctx context.Context, arg DeleteRevokedChannelInstallationByAppIDParams) (pgtype.UUID, error) {
-	row := q.db.QueryRow(ctx, deleteRevokedChannelInstallationByAppID,
-		arg.ChannelType,
-		arg.AppID,
-		arg.WorkspaceID,
-		arg.AgentID,
-	)
-	var id pgtype.UUID
-	err := row.Scan(&id)
-	return id, err
-}
-
 const findReusableChannelUserBinding = `-- name: FindReusableChannelUserBinding :one
 SELECT b.id, b.workspace_id, b.multica_user_id, b.installation_id, b.channel_type, b.channel_user_id, b.config, b.bound_at FROM channel_user_binding b
 JOIN channel_installation ci ON ci.id = b.installation_id
@@ -711,6 +691,42 @@ func (q *Queries) GetChannelInstallationInWorkspace(ctx context.Context, arg Get
 	return i, err
 }

+const getChannelInstallationOwnerByAppID = `-- name: GetChannelInstallationOwnerByAppID :one
+SELECT ci.workspace_id, ci.agent_id, a.archived_at AS agent_archived_at
+FROM channel_installation ci
+JOIN agent a ON a.id = ci.agent_id
+WHERE ci.channel_type = $1
+  AND ci.config ->> 'app_id' = $2::text
+`
+
+type GetChannelInstallationOwnerByAppIDParams struct {
+	ChannelType string `json:"channel_type"`
+	AppID       string `json:"app_id"`
+}
+
+type GetChannelInstallationOwnerByAppIDRow struct {
+	WorkspaceID     pgtype.UUID        `json:"workspace_id"`
+	AgentID         pgtype.UUID        `json:"agent_id"`
+	AgentArchivedAt pgtype.Timestamptz `json:"agent_archived_at"`
+}
+
+// Identifies the LIVE owner of a (channel_type, config->>'app_id') routing slot
+// so the install path can refuse a rebind with an ACCURATE message instead of the
+// old catch-all "connected to a different Multica workspace". Meant to be read
+// only after ReclaimDeadChannelInstallationByAppID has removed every DEAD owner,
+// so a returned row is a live active owner. `agent_archived` distinguishes an
+// archived (reversible) owner — its bot stays owned, recovered by unarchiving the
+// agent or disconnecting the bot — from a plain active one. The JOIN drops a row
+// whose agent no longer exists (an orphan the reclaim gate should already have
+// cleared), so a missing row (pgx.ErrNoRows) means "no live owner". The caller
+// reads agent_archived_at.Valid to tell an archived (reversible) owner apart.
+func (q *Queries) GetChannelInstallationOwnerByAppID(ctx context.Context, arg GetChannelInstallationOwnerByAppIDParams) (GetChannelInstallationOwnerByAppIDRow, error) {
+	row := q.db.QueryRow(ctx, getChannelInstallationOwnerByAppID, arg.ChannelType, arg.AppID)
+	var i GetChannelInstallationOwnerByAppIDRow
+	err := row.Scan(&i.WorkspaceID, &i.AgentID, &i.AgentArchivedAt)
+	return i, err
+}
+
 const getChannelOutboundCardByTask = `-- name: GetChannelOutboundCardByTask :one
 SELECT id, chat_session_id, task_id, channel_type, channel_chat_id, channel_card_message_id, status, last_patched_at, created_at FROM channel_outbound_card_message
 WHERE task_id = $1
@@ -1031,6 +1047,86 @@ func (q *Queries) PurgeExpiredChannelBindingTokens(ctx context.Context, expiresA
 	return err
 }

+const reclaimDeadChannelInstallationByAppID = `-- name: ReclaimDeadChannelInstallationByAppID :one
+WITH dead AS (
+    DELETE FROM channel_installation ci
+    WHERE ci.channel_type = $1
+      AND ci.config ->> 'app_id' = $2::text
+      AND (
+            (ci.status = 'revoked'
+                AND ci.workspace_id = $3
+                AND ci.agent_id <> $4)
+         OR NOT EXISTS (SELECT 1 FROM workspace w WHERE w.id = ci.workspace_id)
+         OR NOT EXISTS (SELECT 1 FROM agent a WHERE a.id = ci.agent_id)
+      )
+    RETURNING ci.id
+),
+cleared_chat_sessions AS (
+    DELETE FROM channel_chat_session_binding
+    WHERE installation_id IN (SELECT id FROM dead)
+),
+cleared_binding_tokens AS (
+    DELETE FROM channel_binding_token
+    WHERE installation_id IN (SELECT id FROM dead)
+),
+cleared_user_bindings AS (
+    DELETE FROM channel_user_binding
+    WHERE installation_id IN (SELECT id FROM dead)
+),
+detached_audit AS (
+    UPDATE channel_inbound_audit SET installation_id = NULL
+    WHERE installation_id IN (SELECT id FROM dead)
+)
+SELECT id FROM dead
+`
+
+type ReclaimDeadChannelInstallationByAppIDParams struct {
+	ChannelType string      `json:"channel_type"`
+	AppID       string      `json:"app_id"`
+	WorkspaceID pgtype.UUID `json:"workspace_id"`
+	AgentID     pgtype.UUID `json:"agent_id"`
+}
+
+// Rebind cleanup gate. Frees the (channel_type, config->>'app_id') routing slot
+// so a valid new agent can (re)bind a bot whose previous owner is DEAD, and, in
+// the same statement, clears every application-owned dependent row of the removed
+// installation (channel_* has no FK/cascade, MUL-3515 §4). Returns the removed id
+// (pgx.ErrNoRows when nothing was dead — a no-op the caller treats as success).
+//
+// "Dead" is exactly one of:
+//  1. a REVOKED placeholder left by a DIFFERENT agent in the SAME workspace
+//     (Disconnect-then-rebind: agent A revoked, the same bot rebinds to agent B);
+//  2. an ORPHAN whose owning workspace OR agent row no longer exists — the
+//     workspace was deleted, or the agent was hard-deleted on runtime teardown.
+//     With no FK the installation outlives its owner and keeps occupying the
+//     app_id slot: the "ghost binding" that made a bot un-rebindable (#4810).
+//
+// Deliberately NOT dead (the caller refuses these with an accurate conflict):
+//   - the SAME agent's own revoked row (agent_id = @agent_id): the upsert
+//     reactivates it in place, preserving its installation_id and every binding;
+//   - a live ACTIVE owner whose agent still exists — INCLUDING an ARCHIVED agent:
+//     archive is reversible, so its bot stays owned rather than being silently
+//     stolen. Only a hard delete frees the slot.
+//
+// The guard lives in the DELETE predicate (not a prior SELECT) so under READ
+// COMMITTED the row is re-checked at execution (EvalPlanQual): a concurrent
+// same-agent reconnect that flips the revoked row back to 'active' first makes
+// the predicate re-check fail, this deletes nothing, and no dependents are
+// touched — closing the read-then-delete TOCTOU. Dependent cleanup keys off the
+// actually-deleted id (the `dead` CTE), so it runs ONLY for a row this statement
+// removed. The (channel_type, app_id) unique index guarantees at most one match.
+func (q *Queries) ReclaimDeadChannelInstallationByAppID(ctx context.Context, arg ReclaimDeadChannelInstallationByAppIDParams) (pgtype.UUID, error) {
+	row := q.db.QueryRow(ctx, reclaimDeadChannelInstallationByAppID,
+		arg.ChannelType,
+		arg.AppID,
+		arg.WorkspaceID,
+		arg.AgentID,
+	)
+	var id pgtype.UUID
+	err := row.Scan(&id)
+	return id, err
+}
+
 const recordChannelInboundDrop = `-- name: RecordChannelInboundDrop :exec

 INSERT INTO channel_inbound_audit (
--- a/server/pkg/db/generated/workspace.sql.go
+++ b/server/pkg/db/generated/workspace.sql.go
@@ -52,12 +52,37 @@ func (q *Queries) CreateWorkspace(ctx context.Context, arg CreateWorkspaceParams
 }

 const deleteWorkspace = `-- name: DeleteWorkspace :exec
-WITH deleted_pending_check_suites AS (
+WITH ws_installations AS (
+    SELECT id FROM channel_installation WHERE workspace_id = $1
+),
+cleared_chat_sessions AS (
+    DELETE FROM channel_chat_session_binding WHERE installation_id IN (SELECT id FROM ws_installations)
+),
+detached_audit AS (
+    UPDATE channel_inbound_audit SET installation_id = NULL WHERE installation_id IN (SELECT id FROM ws_installations)
+),
+cleared_user_bindings AS (
+    DELETE FROM channel_user_binding WHERE workspace_id = $1
+),
+cleared_binding_tokens AS (
+    DELETE FROM channel_binding_token WHERE workspace_id = $1
+),
+cleared_installations AS (
+    DELETE FROM channel_installation WHERE workspace_id = $1
+),
+deleted_pending_check_suites AS (
    DELETE FROM github_pending_check_suite WHERE workspace_id = $1
 )
-DELETE FROM workspace WHERE id = $1
+DELETE FROM workspace WHERE workspace.id = $1
 `

+// The channel_* tables carry NO FK to workspace (MUL-3515 §4), so — unlike the
+// CASCADE-backed tables the DELETE below sweeps — they are not cleaned up
+// implicitly. Remove this workspace's channel installations, and every dependent
+// row of each, here so a deleted workspace never leaves an orphaned installation
+// occupying its bot's (channel_type, config->>'app_id') routing slot, which would
+// make that bot un-rebindable anywhere until an operator hand-deletes the row
+// (#4810). All in one statement so it commits atomically with the workspace row.
 func (q *Queries) DeleteWorkspace(ctx context.Context, id pgtype.UUID) error {
 	_, err := q.db.Exec(ctx, deleteWorkspace, id)
 	return err
--- a/server/pkg/db/queries/channel.sql
+++ b/server/pkg/db/queries/channel.sql
@@ -98,40 +98,111 @@ SELECT * FROM channel_installation
 WHERE channel_type = sqlc.arg('channel_type')
  AND config ->> 'app_id' = sqlc.arg('app_id')::text;

-- name: DeleteRevokedChannelInstallationByAppID :one
-- Hard-delete a REVOKED installation that belongs to a DIFFERENT agent, keyed
-- by its platform app identity, and RETURN the deleted id. When a Feishu/Lark
-- Bot is disconnected from agent A (status → 'revoked') and the same Bot (same
-- app_id) is later bound to a DIFFERENT agent B, agent A's revoked row still
-- occupies the (channel_type, config->>'app_id') unique slot and blocks the
-- UpsertChannelInstallation INSERT for B. Removing that placeholder first lets
-- the upsert create a fresh row for the new agent.
+-- name: GetChannelInstallationOwnerByAppID :one
+-- Identifies the LIVE owner of a (channel_type, config->>'app_id') routing slot
+-- so the install path can refuse a rebind with an ACCURATE message instead of the
+-- old catch-all "connected to a different Multica workspace". Meant to be read
+-- only after ReclaimDeadChannelInstallationByAppID has removed every DEAD owner,
+-- so a returned row is a live active owner. `agent_archived` distinguishes an
+-- archived (reversible) owner — its bot stays owned, recovered by unarchiving the
+-- agent or disconnecting the bot — from a plain active one. The JOIN drops a row
+-- whose agent no longer exists (an orphan the reclaim gate should already have
+-- cleared), so a missing row (pgx.ErrNoRows) means "no live owner". The caller
+-- reads agent_archived_at.Valid to tell an archived (reversible) owner apart.
+SELECT ci.workspace_id, ci.agent_id, a.archived_at AS agent_archived_at
+FROM channel_installation ci
+JOIN agent a ON a.id = ci.agent_id
+WHERE ci.channel_type = sqlc.arg('channel_type')
+  AND ci.config ->> 'app_id' = sqlc.arg('app_id')::text;
+
+-- name: ReclaimDeadChannelInstallationByAppID :one
+-- Rebind cleanup gate. Frees the (channel_type, config->>'app_id') routing slot
+-- so a valid new agent can (re)bind a bot whose previous owner is DEAD, and, in
+-- the same statement, clears every application-owned dependent row of the removed
+-- installation (channel_* has no FK/cascade, MUL-3515 §4). Returns the removed id
+-- (pgx.ErrNoRows when nothing was dead — a no-op the caller treats as success).
 --
-- The `agent_id <> arg` fence is load-bearing: re-connecting the SAME agent
-- must NOT delete its own revoked row. UpsertChannelInstallation conflicts on
-- (workspace_id, agent_id, channel_type), so the same agent's revoked row is
-- reactivated in place (status → 'active') with its installation_id — and every
-- channel_user_binding / channel_chat_session_binding hanging off it —
-- preserved. Deleting it here would force an INSERT with a fresh installation_id
-- and orphan all of that agent's member links and chat sessions.
+-- "Dead" is exactly one of:
+--   1. a REVOKED placeholder left by a DIFFERENT agent in the SAME workspace
+--      (Disconnect-then-rebind: agent A revoked, the same bot rebinds to agent B);
+--   2. an ORPHAN whose owning workspace OR agent row no longer exists — the
+--      workspace was deleted, or the agent was hard-deleted on runtime teardown.
+--      With no FK the installation outlives its owner and keeps occupying the
+--      app_id slot: the "ghost binding" that made a bot un-rebindable (#4810).
 --
-- Fenced to one workspace AND status='revoked' so an active installation can
-- never be silently deleted through this path.
+-- Deliberately NOT dead (the caller refuses these with an accurate conflict):
+--   - the SAME agent's own revoked row (agent_id = @agent_id): the upsert
+--     reactivates it in place, preserving its installation_id and every binding;
+--   - a live ACTIVE owner whose agent still exists — INCLUDING an ARCHIVED agent:
+--     archive is reversible, so its bot stays owned rather than being silently
+--     stolen. Only a hard delete frees the slot.
 --
-- This DELETE is the atomic gate for the whole rebind cleanup: the caller keys
-- its dependent-row cleanup off the RETURNING id, so cleanup runs ONLY for a row
-- this statement actually removed. Under READ COMMITTED, a concurrent same-agent
-- reconnect that reactivates the row to 'active' first makes the WHERE re-check
-- fail (EvalPlanQual), this deletes nothing (pgx.ErrNoRows), and no dependents
-- are touched — closing the read-then-delete TOCTOU where a stale "it was
-- revoked" read could wipe the bindings of a since-reactivated installation.
-DELETE FROM channel_installation
-WHERE channel_type = sqlc.arg('channel_type')
-  AND config ->> 'app_id' = sqlc.arg('app_id')::text
-  AND workspace_id = sqlc.arg('workspace_id')
-  AND agent_id <> sqlc.arg('agent_id')
-  AND status = 'revoked'
-RETURNING id;
+-- The guard lives in the DELETE predicate (not a prior SELECT) so under READ
+-- COMMITTED the row is re-checked at execution (EvalPlanQual): a concurrent
+-- same-agent reconnect that flips the revoked row back to 'active' first makes
+-- the predicate re-check fail, this deletes nothing, and no dependents are
+-- touched — closing the read-then-delete TOCTOU. Dependent cleanup keys off the
+-- actually-deleted id (the `dead` CTE), so it runs ONLY for a row this statement
+-- removed. The (channel_type, app_id) unique index guarantees at most one match.
+WITH dead AS (
+    DELETE FROM channel_installation ci
+    WHERE ci.channel_type = sqlc.arg('channel_type')
+      AND ci.config ->> 'app_id' = sqlc.arg('app_id')::text
+      AND (
+            (ci.status = 'revoked'
+                AND ci.workspace_id = sqlc.arg('workspace_id')
+                AND ci.agent_id <> sqlc.arg('agent_id'))
+         OR NOT EXISTS (SELECT 1 FROM workspace w WHERE w.id = ci.workspace_id)
+         OR NOT EXISTS (SELECT 1 FROM agent a WHERE a.id = ci.agent_id)
+      )
+    RETURNING ci.id
+),
+cleared_chat_sessions AS (
+    DELETE FROM channel_chat_session_binding
+    WHERE installation_id IN (SELECT id FROM dead)
+),
+cleared_binding_tokens AS (
+    DELETE FROM channel_binding_token
+    WHERE installation_id IN (SELECT id FROM dead)
+),
+cleared_user_bindings AS (
+    DELETE FROM channel_user_binding
+    WHERE installation_id IN (SELECT id FROM dead)
+),
+detached_audit AS (
+    UPDATE channel_inbound_audit SET installation_id = NULL
+    WHERE installation_id IN (SELECT id FROM dead)
+)
+SELECT id FROM dead;
+
+-- name: DeleteChannelInstallationsByArchivedRuntimeAgents :exec
+-- Application-layer replacement for the (deliberately absent, MUL-3515 §4)
+-- workspace/agent ON DELETE CASCADE: on runtime teardown, before the archived
+-- agents are hard-deleted, remove every channel installation they own — plus all
+-- of each installation's dependent rows — so no orphaned installation keeps
+-- occupying its bot's (channel_type, app_id) routing slot after its agent is gone
+-- (#4810). MUST run in the same tx as, and BEFORE, DeleteArchivedAgentsByRuntime.
+-- Mirrors the agent hard-delete predicate (runtime_id, archived_at IS NOT NULL)
+-- exactly.
+WITH doomed AS (
+    SELECT id FROM channel_installation
+    WHERE agent_id IN (
+        SELECT id FROM agent WHERE runtime_id = sqlc.arg('runtime_id') AND archived_at IS NOT NULL
+    )
+),
+cleared_chat_sessions AS (
+    DELETE FROM channel_chat_session_binding WHERE installation_id IN (SELECT id FROM doomed)
+),
+cleared_binding_tokens AS (
+    DELETE FROM channel_binding_token WHERE installation_id IN (SELECT id FROM doomed)
+),
+cleared_user_bindings AS (
+    DELETE FROM channel_user_binding WHERE installation_id IN (SELECT id FROM doomed)
+),
+detached_audit AS (
+    UPDATE channel_inbound_audit SET installation_id = NULL WHERE installation_id IN (SELECT id FROM doomed)
+)
+DELETE FROM channel_installation WHERE id IN (SELECT id FROM doomed);

 -- name: ListChannelInstallationsByWorkspace :many
 -- Scoped by channel_type so a per-channel management surface (e.g. the Lark
--- a/server/pkg/db/queries/workspace.sql
+++ b/server/pkg/db/queries/workspace.sql
@@ -46,7 +46,32 @@ WHERE id = $1
 RETURNING issue_counter;

 -- name: DeleteWorkspace :exec
-WITH deleted_pending_check_suites AS (
+-- The channel_* tables carry NO FK to workspace (MUL-3515 §4), so — unlike the
+-- CASCADE-backed tables the DELETE below sweeps — they are not cleaned up
+-- implicitly. Remove this workspace's channel installations, and every dependent
+-- row of each, here so a deleted workspace never leaves an orphaned installation
+-- occupying its bot's (channel_type, config->>'app_id') routing slot, which would
+-- make that bot un-rebindable anywhere until an operator hand-deletes the row
+-- (#4810). All in one statement so it commits atomically with the workspace row.
+WITH ws_installations AS (
+    SELECT id FROM channel_installation WHERE workspace_id = $1
+),
+cleared_chat_sessions AS (
+    DELETE FROM channel_chat_session_binding WHERE installation_id IN (SELECT id FROM ws_installations)
+),
+detached_audit AS (
+    UPDATE channel_inbound_audit SET installation_id = NULL WHERE installation_id IN (SELECT id FROM ws_installations)
+),
+cleared_user_bindings AS (
+    DELETE FROM channel_user_binding WHERE workspace_id = $1
+),
+cleared_binding_tokens AS (
+    DELETE FROM channel_binding_token WHERE workspace_id = $1
+),
+cleared_installations AS (
+    DELETE FROM channel_installation WHERE workspace_id = $1
+),
+deleted_pending_check_suites AS (
    DELETE FROM github_pending_check_suite WHERE workspace_id = $1
 )
-DELETE FROM workspace WHERE id = $1;
+DELETE FROM workspace WHERE workspace.id = $1;