* fix: sync suppression list removal with AWS SES (closes#324)
When removing an email from the suppression list, now also removes it from
AWS SES account-level suppression list across all regions where the team
has domains configured.
- Add deleteFromSesSuppressionList helper to ses.ts
- Update removeSuppression to query team domains for unique regions
- Use best-effort pattern: AWS failures don't block local DB deletion
- Handle NotFoundException gracefully (email not in SES list)
* fix: correct failure detection logic for SES suppression removal
deleteFromSesSuppressionList returns false on error (never throws),
so check for fulfilled promises with value === false instead of
rejected status.
* fix: account for rejected promises in SES suppression removal
Updated the filter logic for Promise.allSettled to include 'rejected'
status as well as 'fulfilled' with a 'false' value. This ensures that
any errors occurring before the try block in deleteFromSesSuppressionList
are correctly caught and logged.
* feat: add web test framework with infra-backed suites
* fix: honor DATABASE_URL env in integration prepare script
* fix: apply web test review feedback
* fix: streamline web test infra lifecycle and workflow scope
The warning and limit-reached notification emails were being sent
multiple times because of a race condition: concurrent workers could
both read the Redis cooldown key as empty (GET), both send emails,
then both set the key (SETEX). Replaced the non-atomic GET + SETEX
pattern with a single atomic SET ... NX EX that claims the cooldown
slot before any emails are sent. Also increased cooldown from 6 hours
to 24 hours so each notification is sent at most once per day.
https://claude.ai/code/session_01VBYXi5e64Vtq1cXWsfTYTw
Co-authored-by: Claude <noreply@anthropic.com>