Document retry behavior for stale inactive reconnect cleanup

Co-authored-by: Eric Allam <eric@trigger.dev>

Author: cursoragent

docs/tasks/streams.mdx

@@ -650,6 +650,7 @@ class MemoryRunStore implements TriggerChatRunStore {
 the transport clears stored run state and future reconnect attempts return `null`.
 If stale inactive reconnect state cannot be cleaned up, reconnect still returns `null` and
 the failure is surfaced through `onError` with phase `reconnect`.
+Subsequent reconnect calls will retry stale inactive-state cleanup until it succeeds.
 
 For richer TypeScript ergonomics in app code, `@trigger.dev/ai` also exports:
 

packages/ai/CHANGELOG.md

@@ -18,3 +18,4 @@
 - Added best-effort run-store cleanup so cleanup failures do not mask root transport errors.
 - Improved best-effort run-store cleanup to attempt both inactive-state writes and deletes even if one step fails.
 - Added reconnect cleanup error reporting for stale inactive state while still returning `null`.
+- Added retry semantics for stale inactive reconnect cleanup on subsequent reconnect attempts.

packages/ai/README.md

@@ -150,6 +150,7 @@ both cleanup steps (`set` inactive state and `delete`) even if one of them fails
 - Once a stream completes or errors, its run state is cleaned up and reconnect returns `null`.
 - If reconnect finds stale inactive state and run-store cleanup fails, `onError` receives a
   `"reconnect"` phase event and reconnect still returns `null`.
+- If inactive-state cleanup fails, later reconnect calls retry that cleanup until it succeeds.
 - Provide a custom `runStore` if you need state shared across processes/instances.
 
 ## `ai.tool(...)` example