01a07197 - Refresh the E2E test image on every stack startup

[TaprootFreakAI]

EN:
Refresh the Playwright test image on every stack startup so repeated runs use the current checked-out tests. Move the existing build from run.sh into up.sh and update the troubleshooting documentation.

DE:
Das Playwright-Test-Image bei jedem Stack-Start aktualisieren, damit wiederholte Läufe die aktuell ausgecheckten Tests verwenden. Der vorhandene Build wird von run.sh nach up.sh verschoben und die Fehlerbehebungs-Dokumentation angepasst.

Cause, change and validation

Compose builds a missing tests image but reuses an existing image even when the checkout has changed. The specs are copied into the image, without bind mounts. A CI path that calls up.sh and then runs tests in a separate shell could therefore silently execute old specs when reusing a Compose project.

Call the existing build_tests_image helper from up.sh after frontend/widget image handling and before service startup. Remove the redundant invocation from run.sh, which already calls up.sh. This also applies when the frontend images are prebuilt; normal Docker build caching remains enabled and a build error still stops startup.

The defect was confirmed by inspecting a reused image: it lacked the two synchronization regressions already present in the checkout, and its run executed 262 rather than 264 cases. This follow-up prevents that stale-image path at its shared entry point.

Validation: full GitHub-hosted CI 34027975470 passed at 7da6991. The log proves the tests-image build inside up.sh and 264 E2E cases: 256 ordinary passes, 5 expected failures and 3 skips, with both synchronization regressions passing. All 1,828 unit tests, lint, Markdown, application/widget builds and harness type-check passed; CodeQL and the configured review workflow also succeeded. No local App CI is run, as requested. No production code, dependency, workflow or test-assertion changes. Existing expected failures and skipped cases remain explicit.

Independent final reviews were explicitly waived by the requester: 0 independent final-review approvals are claimed. The preceding PR #1451 is already merged, so this correction is a separate follow-up PR.

[TaprootFreakAI]

EN:
Independent final reviews were explicitly waived by the requester; 0 independent review approvals are claimed.
Rebuild the Playwright test image on every stack startup so separate CI shells and repeated projects run current tests.

DE:
Die unabhängigen Abschlussreviews wurden ausdrücklich vom Auftraggeber erlassen; es werden 0 unabhängige Review-Freigaben beansprucht.
Das Playwright-Test-Image bei jedem Stack-Start neu bauen, damit getrennte CI-Aufrufe und wiederverwendete Projekte die aktuellen Tests ausführen.

Validation

Head: 7da6991; base: 3d8045d. Signed commit verified by GitHub. Full hosted CI 34027975470 succeeded: 1,828 unit tests; 264 E2E cases (256 ordinary passes, 5 expected failures, 3 skips), both synchronization regressions passed. The startup log confirms rebuilding the test image before service startup. Lint, Markdown, application/widget builds, harness type-check, CodeQL and configured review workflow succeeded. No local App CI was run.

All four comment sources were fetched with pagination: no issue comments, reviews, inline comments or review threads. Mergeability is MERGEABLE. No required status checks in the active branch rules. Historical label-gated skips are superseded by the successful ci:full runs. No production code, dependencies or assertions changed.