Ian Rose

> It looks like this also closes https://github.com/dask/dask/issues/8280? I haven't really tackled that issue here -- it seems more involved and may require actual work on the algorithm (based on...

Just added one more commit adding some additional test coverage, in case you have a few minutes @jrbourbeau

Thanks for the xref @jorloplaz, I agree they look the same. As more and more pandas custom dtypes are added/promoted-from-experimental, this is going to be more important.

> > > Any more thoughts on this? > > > > > > I still think what I said here [#8750 (comment)](https://github.com/dask/dask/pull/8750#issuecomment-1095251068), that we shouldn't introduce a new namespace...

I like contracts :)

I ran into this issue again yesterday. It turns out to be a pretty big problem when trying to use sparse arrays with dask, in that any kind of reduction...

As @pavithraes notes, this has been happening with regularity on python3.10+ubuntu for the last week+. I see there was a new release of h5py that happened around the same time....

Yes, I was about to open a PR testing that. I'm concerned that it came up at all, though, in the context of using a dask lock. It seems that...

Fair enough -- I have opened #9154, if that seems to improve things, I'm happy to move on

I'll keep an eye on it tomorrow, and if things seem better by EOD, I can close it then