We bought a company hosting on Azure. They used hosted Postgres and are hosting .NET services on Windows. Small infra, in range of 2-3 hundreds cores and 1T mem

We bought a company hosting on Azure. They used hosted Postgres and are hosting .NET services on Windows. Small infra, in range of 2-3 hundreds cores and 1T memory. Every few days M$ randomly shuts down random instances for maintenance, disconnects network for >10 minutes. Migrated off hosted Postgres because performance was tragedy - now their India-based expert led us to use different volumes type and after instance restart database didn’t start up because of I/O latency. Expert don’t want to meet for 3 straight days now, because he is busy. RCA (half pager, written probably by some LLM) says it’s not their fault, but charts says different story. The only thing they crash GCP and AWS with is dashboard that loads everything so quickly... sad you can’t run e.g. making 2 similar network operations in parallel because they will fail or take 10x the time they would take when run one after another. Run, don’t use.