Huite

The inconsistent chunks makes sense right? Because you're opening the datasets one by one, xarray/dask will pick a chunksize per dataset. Since the datasets are differently sized, the chunks are...

chunks="auto" should work fine, but resulted in inconsistent chunks error. This will be fixed by: #280 After merging the PR, using "auto" will give reasonable performance.

This is a familiar problem to me, I ran into the same issues setting up cyclical forcing for MODFLOW off of date times. In that particular case, I was hoping...

I think this is fixed in a14e4501925831754628fd774661ed0895908e8b, but ideally we do a check.

Good to know: the problem here, I presume is that my wrapping is once again incomplete. Attribute and method access does a check whether a xugrid object should be returned....

I must admit I haven't looked at the arrow input much, but this: `model.basin.profile.set_filepath(Path("profile.arrow"))` is not very intuitive. I would prefer a property, `basin.to_arrow = True` or something. Ideally, for...

> I'd hold off on write(arrow=True) since the downsides are currently no QGIS support https://github.com/Deltares/Ribasim/issues/318, and these files are usually not editable. You can keep the default value `False`. For...

I think my current recommendation would be to replace it by a DarcyWeisbachResistance (or something) instead, since it's derived on more fundamental grounds rather than empiric Manning or Chezy. However,...

Do you mean starting the relaxation at Re < 3000 (approx. laminar flow) instead of the current arbitrary threshold of 1e-5? Quick note, if you check the wikipedia Moody diagram...

> My proposal of relaxing based on Re < 2000 makes the model incredibly slow. I have benchmarked the model HollandseDelta_parameterized_2025_7_1, which runs in the old implementation in 6 minutes,...