Ok, this is a good point, but some clarifications first:

• TimescaleDB gives time series database capabilities to Postgres.
• We don't use Prometheus for app metrics
• With the currently planned architecture, we app metrics service talks to Postgres directly: no Prometheus.

It is possible that we don't use Timescaledb. In this case, we will just have regular SQL. There will be performance problems eventually because we won't use any time optimal things that TimescaleDB provides.

I honestly didn't think about not using TimescaleDB at all since my research and previous experience has told me that we would need something optimal for time series data.

It will definitely work for the very initial MVP though.