TypeError: Invalid value for attr '_FillValue': b''.

I'm trying to combine three netcdf file jsons using the following notebook:
https://nbviewer.org/gist/6117f63ba1b266ea713249b98492ce76
(which could be reproduced if you have AWS credentials -- it reads from a requester pays bucket)

I'm getting this error, which I really don't understand, since we are not creating NetCDF files and the original NetCDF files don't have any fill values.

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
File <timed exec>:2, in <module>

File /home/conda/store/26851070ae96f50ac395255946c7876bb86f109ca9ae22af854d21e68393d114-pangeo/lib/python3.9/site-packages/kerchunk/combine.py:70, in MultiZarrToZarr.translate(self, outpath, template_count)
     55 """
     56 Translate the combined reference files and write to new file
     57 
   (...)
     65 
     66 """
     69 ds, ds0, fss = self._determine_dims()
---> 70 out = self._build_output(ds, ds0, fss)
     71 self.output = self._consolidate(out, template_count=template_count)
     73 if outpath:

File /home/conda/store/26851070ae96f50ac395255946c7876bb86f109ca9ae22af854d21e68393d114-pangeo/lib/python3.9/site-packages/kerchunk/combine.py:171, in MultiZarrToZarr._build_output(self, ds, ds0, fss)
    169 out = {}
    170 logger.debug("write zarr metadata")
--> 171 ds.to_zarr(out, chunk_store={}, compute=False,
    172            consolidated=False)  # fills in metadata&coords
    173 z = zarr.open_group(out, mode='a')
    174 accum_dim = list(self.concat_dims.union(self.extra_dims))[0]  # only ever one dim for now

File /home/conda/store/26851070ae96f50ac395255946c7876bb86f109ca9ae22af854d21e68393d114-pangeo/lib/python3.9/site-packages/xarray/core/dataset.py:2035, in Dataset.to_zarr(self, store, chunk_store, mode, synchronizer, group, encoding, compute, consolidated, append_dim, region, safe_chunks, storage_options)
   2032 if encoding is None:
   2033     encoding = {}
-> 2035 return to_zarr(
   2036     self,
   2037     store=store,
   2038     chunk_store=chunk_store,
   2039     storage_options=storage_options,
   2040     mode=mode,
   2041     synchronizer=synchronizer,
   2042     group=group,
   2043     encoding=encoding,
   2044     compute=compute,
   2045     consolidated=consolidated,
   2046     append_dim=append_dim,
   2047     region=region,
   2048     safe_chunks=safe_chunks,
   2049 )

File /home/conda/store/26851070ae96f50ac395255946c7876bb86f109ca9ae22af854d21e68393d114-pangeo/lib/python3.9/site-packages/xarray/backends/api.py:1375, in to_zarr(dataset, store, chunk_store, mode, synchronizer, group, encoding, compute, consolidated, append_dim, region, safe_chunks, storage_options)
   1373 # validate Dataset keys, DataArray names, and attr keys/values
   1374 _validate_dataset_names(dataset)
-> 1375 _validate_attrs(dataset)
   1377 if region is not None:
   1378     _validate_region(dataset, region)

File /home/conda/store/26851070ae96f50ac395255946c7876bb86f109ca9ae22af854d21e68393d114-pangeo/lib/python3.9/site-packages/xarray/backends/api.py:175, in _validate_attrs(dataset, invalid_netcdf)
    173 for variable in dataset.variables.values():
    174     for k, v in variable.attrs.items():
--> 175         check_attr(k, v, valid_types)

File /home/conda/store/26851070ae96f50ac395255946c7876bb86f109ca9ae22af854d21e68393d114-pangeo/lib/python3.9/site-packages/xarray/backends/api.py:162, in _validate_attrs.<locals>.check_attr(name, value, valid_types)
    156     raise TypeError(
    157         f"Invalid name for attr: {name!r} must be a string for "
    158         "serialization to netCDF files"
    159     )
    161 if not isinstance(value, valid_types):
--> 162     raise TypeError(
    163         f"Invalid value for attr {name!r}: {value!r}. For serialization to "
    164         "netCDF files, its value must be of one of the following types: "
    165         f"{', '.join([vtype.__name__ for vtype in valid_types])}"
    166     )

TypeError: Invalid value for attr '_FillValue': b''. For serialization to netCDF files, its value must be of one of the following types: str, Number, ndarray, number, list, tuple

Reproduce the failure with the linked notebook and inspect MultiZarrToZarr.translate and _build_output in kerchunk/combine.py, where the traceback reaches xarray Dataset.to_zarr. Check how the combined dataset metadata reaches xarray's attribute validation; done means the three reference JSONs combine without the reported TypeError.