Every engagement starts the same way, regardless of stack: a short, structured review before we recommend a single change. A few things we always check, in roughly this order.
Who can deploy, and what happens if that person is unreachable. Whether backups have ever been restored. Whether monitoring alerts actually get acted on, or quietly ignored. What's running that nobody currently at the company remembers setting up. And what the last three incidents had in common.
That last one is usually the most useful. Teams rarely have one big infrastructure problem — they have one root cause showing up as three or four different symptoms. Finding the actual pattern, instead of patching each symptom separately, is most of the value of a review before any work even starts.