[indent]

Production logs show one account stuck retrying one ChatGPT attachment. Each of its five real attempts ended with "attachment download interrupted" after about 32 s and logged "Event handling is taking long" for that room. In between, the media retry schedule kept the attachment pending and backed off toward 15 minutes. The file never arrived. Later messages in the room were still delivered.

Cause: incomingFile gives the download a 2-minute context, but DownloadAttachment fetches the storage URL with the shared API client, whose http.Client.Timeout is 30 s. That timeout also covers reading the body, so io.ReadAll fails at 30 s and returns an AttachmentStorageError. Any file that needs more than 30 s to fetch fails the same way on every retry, forever.

Change: the storage GET uses a copy of the client with Timeout set to 0, so the caller's 2-minute context sets the limit. Metadata and other API calls keep the 30 s timeout. go build ./... and go vet pass.

Tag @indent to continue the conversation here.

[indent]

Issues

No issues found.

CI Checks

All CI checks passed on 43113fe.

Review agents

Select any unchecked box below to run or rerun that agent.

Passed (1)

• Indent Review Agent · No issues found in the timeout change.

Full results

Indent Review Agent

• Summary: No issues found in the timeout change.
• Last ran on commit: 43113fe2
• Latest result

  {
    "summary": "No issues found in the timeout change.",
    "findings": []
  }