I think I must have just been remembering the PyTensor announcement and mistakenly thought some of these features were already present:
One example is Xarray-like dimension support:
Xarrayadds many useful features to arrays. Most notably, named dimensions and coordinates. This is what allowscoordsanddimsin PyMC to give names to your arrays. Currently, Aesara does not support this, nor are there plans to do so. But there are many benefits to having labeled tensors as well.
Any issues or discussions to follow on that? I tried searching both the PyTensor and PyMC github repos but came up short.