Follow-up from "fix: ensure DB LB replica resolution timeout does not prevent app startup"

The following discussion from !1876 (merged) should be addressed:

• @hswimelar started a discussion: (+1 comment)

  Technically, we could make this far more complicated by having multiple timeouts for each operation (one global timeout; one timeout for the SRV lookup; one timeout for the Host lookup per replica; one timeout for the DB connection per replica), but that would make this far more complicated to debug and test.

  We should also think about resolving hosts in parallel. With the current setup if the first host times out, we wouldn't try the subsequent ones.