- Shipped
- August 6, 2026 at 4:09 PM UTC
- Author
- Kamo
- Commit
- 949a204
Plan on the request thread (two cheap queries, so the response is exact about who is included and who was skipped and why), write off it: 200 members x 8 documents is 1,600 rows plus as many ESigService calls, and DocsService serves WOPI. The pool is NOT a Spring Executor bean. applicationTaskExecutor is **************** and DocsService already has bare @Async on KSessionService and DocumentService -- publishing an Executor bean would silently move the WOPI session touch onto this small bounded pool. Envelopes are created outside any transaction; a 20s read timeout while holding a Hikari connection is how a pool of ten starves. A signable document whose envelope failed keeps a null envelope_uid, which is the only signal the grid has. The redeliver/void/assignment-grid mappings land here too, so the controller has ONE authorization funnel written and proved once rather than a second one bolted on later; their services answer 501 until Tasks 4-6 fill the bodies.