fix(mcp): acquire a fresh cache mutation per discovery retry

A retried discovery took mutation ownership once before the retry loop, so a
retry that succeeded after a concurrent clearCache published under a stale
ownership id, lost the CAS, and dropped otherwise-valid tools (empty
fail-closed result). Begin a fresh mutation per attempt so a retried result
publishes under a current id. Adds a regression test.

Author: waleedlatif1

apps/sim/lib/mcp/service.test.ts

@@ -517,6 +517,23 @@ describe('McpService.discoverTools per-server caching', () => {
     )
   })
 
+  it('acquires a fresh mutation on each discovery retry so a retried result still publishes', async () => {
+    const serverKey = `workspace:${WORKSPACE_ID}:server:mcp-a`
+    mockGetWorkspaceServersRows.mockResolvedValue([dbRow('mcp-a', 'A')])
+    mockCacheAdapter.beginMutation.mockClear()
+    mockListTools
+      .mockRejectedValueOnce(new Error('Request timed out'))
+      .mockResolvedValueOnce([tool('a1', 'mcp-a')])
+
+    const tools = await mcpService.discoverServerTools(USER_ID, 'mcp-a', WORKSPACE_ID, true)
+
+    expect(tools).toEqual([tool('a1', 'mcp-a')])
+    expect(cacheStore.get(serverKey)?.tools).toEqual([tool('a1', 'mcp-a')])
+    // One begin per attempt: the retry publishes under a current ownership id
+    // instead of a stale pre-loop id that a concurrent clearCache could supersede.
+    expect(mockCacheAdapter.beginMutation).toHaveBeenCalledTimes(2)
+  })
+
   it('keeps an older ordered publisher from superseding a retry-acquired mutation', async () => {
     const serverKey = `workspace:${WORKSPACE_ID}:server:mcp-a`
     mockGetWorkspaceServersRows.mockResolvedValue([dbRow('mcp-a', 'A')])

apps/sim/lib/mcp/service.ts

@@ -1069,10 +1069,12 @@ class McpService {
       }
     }
 
-    const mutation = await this.beginServerCacheMutation(workspaceId, serverId)
-
     for (let attempt = 0; attempt < maxRetries; attempt++) {
       let config: McpServerConfig | null = null
+      // Begin a fresh mutation per attempt. A retry that succeeds after a
+      // concurrent clearCache must publish under a current ownership id — a
+      // stale pre-loop id would lose the CAS and drop otherwise-valid tools.
+      const mutation = await this.beginServerCacheMutation(workspaceId, serverId)
       try {
         logger.info(
           `[${requestId}] Discovering tools from server ${serverId} for user ${userId}${attempt > 0 ? ` (attempt ${attempt + 1})` : ''}`