Мониторинг ошибок для малых команд

the rollback-you-can-trust-at-2am point is the real one, but the part that gets skipped even more is knowing you need it in the first place. most small teams find out about a bad deploy from a user emailing support, not from a system telling them. no error tracking wired up, no alert on the new 500 rate, nothing watching whether the health check is actually meaningful vs just returning 200 because the process is technically alive. the deploy pipeline can be flawless and you still ship a broken release for six hours because nothing was looking. if you're building for this, I'd rate "tells you within minutes that the new version is worse than the old one" above almost anything in the pipeline itself - most teams don't have that until something expensive breaks and they retrofit it after.