Long-term API consistency issues (probably for a 2.0)?

When looking through https://github.com/mne-tools/mne-python/issues/13634, it occurred to me that every once in a while we are reminded of long-running API or consistency issues that pop up from time to time. A couple come to mind:

• get_data vs __getitem__ and use_rounding, whether a last sample is included or not, etc. -- for example https://github.com/mne-tools/mne-python/issues/13634
• picks=None meaning different combinations of (all vs data) x (exclude=() vs exclude="bads") in multiple places

... and I'm sure there are others I'm forgetting (happy to edit this comment based on discussions below if needed).

We could consider having a longer-term clean-up of things like this. I think it would make sense to plan these for a 2.0 release. Packages like pandas and NumPy I think have made strides in this direction cleaning up some long-standing issues in their code using major releases -- we could consider doing the same.

Start by reading issue 13634 and comparing the documented behavior of get_data, getitem, use_rounding, and picks=None across the affected APIs. Identify the inconsistencies and their impact, then review the discussion to determine whether a scoped 2.0 cleanup plan has been agreed. Done means an explicitly defined, maintainable set of API changes rather than a general list of concerns.