CI is pretty inefficient building, testing containers #234
Description
CI is inefficient at building, testing containers. I'm thinking beyond #207. We build containers for testing and then again to push to edge or a new release. It's not just that this takes up time and resources but dodola builds are non-deterministic (because we're not using conda-lock, see #197). We could have a container pass CI tests and then get (re)built and pushed for a release in an untested broken package.
It would be nice to sit and fix but I wonder if it's worth the added trouble and complexity given that this is such a small and purpose-built application.