Nightly is red

[github-actions]

The Nightly workflow failed.

Failing jobs:

• Landing-Feedback: failure
  Run: https://github.com/DevinoSolutions/upup/actions/runs/33731175756

[github-actions]

The Nightly workflow failed.

Failing jobs:

• Drive-Sandbox: failure
• Landing-Feedback: failure
  Run: https://github.com/DevinoSolutions/upup/actions/runs/33851127890

[github-actions]

The Nightly workflow failed.

Failing jobs:

• Drive-Sandbox: failure
• Landing-Feedback: failure
  Run: https://github.com/DevinoSolutions/upup/actions/runs/33953164395

[github-actions]

The Nightly workflow failed.

Failing jobs:

• E2E-Full: failure
• Smoke-Packages: failure
• Drive-Sandbox: failure
• Landing-Feedback: failure
  Run: https://github.com/DevinoSolutions/upup/actions/runs/34020365186

[github-actions]

The Nightly workflow failed.

Failing jobs:

• E2E-Full: failure
• Smoke-Packages: failure
• Landing-Feedback: failure
  Run: https://github.com/DevinoSolutions/upup/actions/runs/34099773526

[github-actions]

The Nightly workflow failed.

Failing jobs:

• Smoke-Packages: failure
  Run: https://github.com/DevinoSolutions/upup/actions/runs/34202513651

[github-actions]

The Nightly workflow failed.

Failing jobs:

• Smoke-Packages: failure
• Landing-Feedback: failure
  Run: https://github.com/DevinoSolutions/upup/actions/runs/34327456489

[AminDhouib]

Triage 2026-09-09: nightly has been red on every run since 2026-08-02, and the failing job rotated night to night, which hid four independent causes. All four are fixed in #378 (merged to dev), promotion to master follows.

Job	Nights (of 9 checked)	Root cause	Fix
Landing feedback e2e	8	PostHog Query API served a cached empty result to the identical-text expect.poll for the whole 90 s; PostHog created_at proves the events were ingested 1-2 s after capture	refresh: 'force_blocking' on every query
Cloud-drive live	6	one-drive download hit vitest's 5 s default timeout	30 s test/hook timeout for a live cloud suite
Package smoke	3 (+ PR #377)	libheif-js 1.23.2 (2026-09-05) resolved from core's ^1.19.8 range grew the opt-in HEIC chunk to 1939 KiB	opt-in chunk budget re-based 1600→2100 KiB
Full E2E + a11y	2	resume-harness ports 53061/53062 sit inside Linux's ephemeral source-port range → EADDRINUSE	ports moved to 31061/31062

Closing this once the first nightly after the fixes is green (a manual workflow_dispatch on dev is running now).

[AminDhouib]

Fixed by #378, promoted to master in #379. Proof: nightly.yml run 34346667864 (manual dispatch on dev with the identical fixes) is green on all eight jobs. If tomorrow's scheduled run on master is red again, the tracker will re-file.