Julius Busecke

Ok I deployed this successful across a subset of datasets for a specific publication [gist](https://gist.github.com/jbusecke/65d0e160ed1d90dac903c9c3f36cf9da) - the code from above is slightly modified to use python async libraries). Several Observations:...

I have dug as deep as I understand in #39, but do not see a quick solution for this. Should we actually remove the threadsafe from the skin wrapper for...

We have decided to shelf this issue for now. In #40 the `threadsafe` was removed from the `skin` logic. It now works but is still slow on a threaded scheduler....

Good points @dcherian. I personally believe that we should have an internal auto-guessing that updates the inferrer kwargs: ```python kwargs = parse_cf(ds) try: kwargs.update(parse_comodo) try: kwargs.update(parse_sgrid) ``` That way anything...

Thanks for this detailed write up. Very helpful for the discussion. I think overall you are very close to what I have in mind. I think we should do the...

I think that goes along the lines of what was discussed today with @jatkinson1000 @TomNicholas and @raehik.

> For _pad_basic this is fine, but for face-connected grids its the implementation is a bit more complicated because pad_axes is not known until after you know you have a...

@TomNicholas I am starting to get back into hopefully more regular maintainer duties. Do you have any time soon to check in on this? Happy to meet up and figure...

Hey @jaard, good to hear from you! I think generally there is nothing that would prevent this from the technical side (the target array can just be broadcasted in the...

Thanks for tracking this @norlandrhagen. I have not 100% convinced myself, but I believe that for the CMIP case at least I am still getting more datasets to pass (without...