Throw Error instead of warning when running tests

I see in the logs many warnings, which I think shouldn't be there, which means we are ignoring problems without noticing in the CI.

I am wondering if we should replace the warning when running tests with a function that throws errors so that our tests can't ignore them.

As a bonus we won't ned to rewire in so many places, but instead use expect for catching.

No files, tests, or entry points are named. Start by locating the warnings emitted while tests run and review where test setup currently rewires them; compare that with the existing expect-based error handling. Done means test runs fail on problems that currently produce warnings, without unnecessary rewiring.

Written by the indexing model from the issue text.