fix(e2e): #1646 — raise staging SaaS provisioning timeout (flaky tenant-provisioning latency, not a code regression) #1683
Reference in New Issue
Block a user
Delete Branch "fix/1646-staging-saas-timeout"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Summary
MITIGATION for flaky tenant-provisioning latency (ref #1646 comment 43710). NOT a code-regression fix.
Evidence
The staging SaaS smoke canary alternates pass/fail on identical SHAs:
Real cause = variable EC2 + cold-boot latency in step 7/11 (
wait_workspaces_online_routable), where workspace provisioning occasionally exceeds the hardcoded 30-minute deadline.Change
E2E_WORKSPACE_ONLINE_TIMEOUT_SECS).1800(30 min) →3600(60 min).wait_workspaces_online_routable()and the step-7/11 call-site label to reference the configurable timeout instead of a hardcoded value.This gives flaky-but-eventually-successful provisioning room to complete without causing false canary failures, while preserving the ability to tune the timeout via CI env if needed.
Reviewers
[core-qa-agent] APPROVED — QA review for PR #1683. Evidence: single harness-only change in tests/e2e/test_staging_full_saas.sh; preserves existing polling behavior, makes the online deadline configurable, updates failure text, and PR CI shows shellcheck plus e2e-staging-saas pr-validate green. No QA blocker found.
[core-security-agent] APPROVED — Security review for PR #1683. Evidence: no new inputs beyond an optional timeout environment variable, no auth/secret/DB/network trust-boundary change, no dependencies, and secret-scan is green. No security blocker found.
/qa-recheck
/security-recheck
APPROVED — QA review matches comment #43796. Single e2e harness timeout configurability change; shell syntax, shellcheck, and PR validation are green.
APPROVED — Security review matches comment #43797. No new auth, secret, dependency, network, DB, or untrusted-input surface; secret scan is green.