[CI] Nightly publishing wrong file...?

I just tried downloading bragg_edge_iron_normalized image from the nightly release assets: https://github.com/scipp/ess/releases/tag/reduced_data_nightly

But the file that was published 3 hours ago is different from what I have from local test.
The latter one is correct with the 1D pixel offset coordinates.
The release assets look like it's still using the previous one...??
Or is it just cacheing issue in github actions and will be gone in a week or sth....?

From Release Assets

pixi exec -s rich -s scitiff scitiff-show-metadata ~/Downloads/bragg_edge_iron_normalized_16x16x256.tiff  --max-depth 6
{
│   'scitiffmeta': {
│   │   'image': {
│   │   │   'data': {'unit': 'dimensionless', 'dtype': 'float32', 'dims': ['t', 'z', 'c', 'y', 'x'], 'shape': [256, 1, 2, 16, 16]},
│   │   │   'masks': {},
│   │   │   'coords': {
│   │   │   │   'y_pixel_offset': {'unit': 'm', 'dtype': 'float32', 'dims': ['y', 'x'], 'shape': [16, 16], 'values': [[...], ..., [...]]},
│   │   │   │   'Ltotal': {'unit': 'm', 'dtype': 'float64', 'dims': ['y', 'x'], 'shape': [16, 16], 'values': [[...], ..., [...]]},
│   │   │   │   'x_pixel_offset': {'unit': 'm', 'dtype': 'float32', 'dims': ['y', 'x'], 'shape': [16, 16], 'values': [[...], ..., [...]]},
│   │   │   │   'tof': {'unit': 'µs', 'dtype': 'float64', 'dims': ['t'], 'shape': [257], 'values': [13000.0, ..., 150000.0]},
│   │   │   │   'c': {'unit': None, 'dtype': 'string', 'dims': ['c'], 'shape': [2], 'values': ['intensities', 'stdevs']}
│   │   │   },
│   │   │   'name': None
│   │   },
│   │   'daq': {'facility': 'Unknown', 'instrument': 'Unknown', 'detector_type': 'Unknown', 'source_type': 'Unknown', 'source': None, 'simulated': None},
│   │   'extra': None,
│   │   'schema_version': '26.6.0'
│   }
}

From Local Testing

pixi exec -s scitiff -s rich scitiff-show-metadata /tmp/pytest-of-sunyoungyoo/pytest-150/test_publish_reduced_scitiff0/bragg_edge_iron_normalized_16x16x256.tiff --max-depth 6
{
│   'scitiffmeta': {
│   │   'image': {
│   │   │   'data': {'unit': 'dimensionless', 'dtype': 'float32', 'dims': ['t', 'z', 'c', 'y', 'x'], 'shape': [256, 1, 2, 64, 64]},
│   │   │   'masks': {},
│   │   │   'coords': {
│   │   │   │   'y_pixel_offset': {'unit': 'm', 'dtype': 'float32', 'dims': ['y'], 'shape': [64], 'values': [-0.006930000148713589, ..., 0.006930000148713589]},
│   │   │   │   'x_pixel_offset': {'unit': 'm', 'dtype': 'float32', 'dims': ['x'], 'shape': [64], 'values': [-0.006930000148713589, ..., 0.006930000148713589]},
│   │   │   │   'Ltotal': {'unit': 'm', 'dtype': 'float64', 'dims': ['x', 'y'], 'shape': [64, 64], 'values': [[...], ..., [...]]},
│   │   │   │   'tof': {'unit': 'µs', 'dtype': 'float64', 'dims': ['t'], 'shape': [257], 'values': [13000.0, ..., 150000.0]},
│   │   │   │   'c': {'unit': None, 'dtype': 'string', 'dims': ['c'], 'shape': [2], 'values': ['intensities', 'stdevs']}
│   │   │   },
│   │   │   'name': None
│   │   },
│   │   'daq': {'facility': 'Unknown', 'instrument': 'Unknown', 'detector_type': 'Unknown', 'source_type': 'Unknown', 'source': None, 'simulated': None},
│   │   'extra': None,
│   │   'schema_version': '26.6.0'
│   }
}

Start by comparing the nightly release asset at reduced_data_nightly with the local pytest output using scitiff-show-metadata, focusing on the image shape and pixel-offset coordinate dimensions. Trace the nightly publishing workflow and its artifact or cache inputs; done means the published TIFF matches the expected local output, including the 64×64 data and 1D x/y pixel offsets.