Neuroglancer appears to be having difficulty decompressing a Zstd compressed dataset:
https://neuroglancer-demo.appspot.com/#!https://s3proxy.janelia.org/shroff-public/shroff_20220523_Janelia_embryo2/20220523i_Janelia_embryo2_OTO_6x6x6nm_2xbin.zstd.zarr/s6/ng.json

The Google Chrome developer tools console reports the following error:

Error retrieving chunk [object Object]:0,0,0: Error: Invalid typed array length: 4294967226

Raw or Gzip compressed Zarr v2 datasets seem to load just fine:

Raw compressed Zarr v2 dataset:
https://neuroglancer-demo.appspot.com/#!https://s3proxy.janelia.org/shroff-public/shroff_20220523_Janelia_embryo2/20220523i_Janelia_embryo2_OTO_6x6x6nm_2xbin.raw.zarr/s6/ng.json

Gzip compressed Zarr v2 dataset:
https://neuroglancer-demo.appspot.com/#!https://s3proxy.janelia.org/shroff-public/shroff_20220523_Janelia_embryo2/20220523i_Janelia_embryo2_OTO_6x6x6nm_2xbin.gzip.zarr/s6/ng.json

The value 4294967226 corresponds to the 32-bit value 0xFFFFFFBA. The 64-bit value 0xFFFFFFFFFFFFFFBA corresponds to the Zstd error code for "Destination buffer is too small".

The dataset was encoded by Tensorstore 0.1.62 via conda-forge package with build number py312h7e2185d_0 and zstd version 1.5.6 ha6fb4c9_0