Eileen

> maybe even disable transactional fixtures for system tests We wouldn't be able to do this without building in functionality like Database Cleaner. It was an intentional design decision to...

With the current implementation (which I'm working on improving) you are required to set a default shard. Without being able to see the actual code or an example of the...

> Any clue? If you can make a demo application that reproduces this (without oracle) I can try to figure out what changed, but I don't know offhand and can't...

Ok I've identified the source of the issue but I don't know the fix because it wasn't caused by a change to the database connections but rather a change to...

Ok I have figured out what's going on here, why the change had an impact, but I can't fix it, easily because there is no obvious path to a fix....

> But this at the cost of greater complexity. I don't understand how this increases complexity, it's should only require changing the name and callers of `fr` to `default`.

If you're using the ActiveRecord::Base connection as your router to find the tenants then you could set up all your connections in `ApplicationRecord` instead and delete `ComdevRecord`: ```ruby class ApplicationRecord...

> Would you accept a pull request to add this fallback to Type.adapter_name_from in order to make it independent of the presence of the default_shard? No, I don't think a...

Hey @ziggythehamster - I'm having a hard time following the use case here and the full reason `none?` is inadequate for this problem. Can you make an app or script...

I'm not offended you didn't cc me, I still saw the email about the test but didn't merge it because I wanted to wait if Matthew had any objections or...