Tim Hoffmann

Thanks for the pointer I was not aware of this. I suppose we agree the current situation is not ideal. We can reduce the nav height in two ways: (1)...

> But I can't help but think that the right path forward is not a bunch of tiny incremental changes that will push and pull the header in directions that...

Thanks. A better scaling would be much appreciated, as many projects have a lot of content in the nav (e.g. [matplotlib](https://matplotlib.org/devdocs/plot_types/index.html), [numpy](https://numpy.org/doc/stable/reference/index.html), [scipy](https://docs.scipy.org/doc/scipy/)). But I'm afraid I've already contributed all...

I've first noted this on an iPad in landscale orientation - not sure how often such docs is viewed there, but you cannot avoid it and the hori. It also...

> [@konstantint](https://github.com/konstantint) IMO, instead of relying on pyplot as a fallback and a concept of a global "active plot", you should instead refactor things so that you explicitly create and...

Just to be clear: I think the original issue here and https://github.com/konstantint/matplotlib-venn/issues/87#issuecomment-3429254593 are unrelated. The latter is a result of parallelizing pytest and simultaneously using a GUI backend. See https://github.com/matplotlib/matplotlib/issues/30666#issuecomment-3429765198

`clear()` semantics for shared Axes is not well defined. See also https://github.com/matplotlib/matplotlib/issues/28851#issuecomment-2371603978. As an implementation detail, sharing is implemented asymmetrically. ```python In [1]: fig, axs = plt.subplots(1, 2, sharey=True) In...

I think a reasonable incremental improvement would be to remove the `Axes._sharex/y` state. It's only [used in very few places](https://github.com/search?q=repo%3Amatplotlib/matplotlib%20_sharex&type=code) and can more reasonably be handled via the grouper `Axes._shared_axes['x']`.

If possible, let's also backport to 3.9 to have a version that supports both the old and new syntax and thus have a smoother transition.

Not necessarily to be done in this PR, but the docstring should be substantially enhanced to explain what this is for and how it works, some points are: - you...