In #1222, it was observed that a new supported file like .coveragerc.toml can put us in a tricky spot with respect to defaults. In theory, it would be good to stop recommending .coveragerc and prefer the TOML format.

But in practice, this is messy because older versions of Coverage.py wouldn't support the new format. So we'd need some sort of adaptive logic, the complexity of which is obviously not advisable (e.g. where do we look to determine versions? The runtime? venv? project deps?)

But then we're sort-of paralyzed, not able to ever change recommendations. I suppose you can just have a gradual period of ecosystem shift but especially with dev config it often sits there for years without being bumped so I foresee some issues with that approach. And it feels unprincipled.

It would be good to give this some thought