The delay to mark the database as unhealthy should be configurable. currently, it looks like hardcoded to 15 milliseconds.

I have a deployment where the remote db has 28-29 ms latency, and Coder is marking it as unhealthy.