This started failing with chexus==24.12.0 as it had #39

If I remember correctly the conclusion on Slack was that this is allowed if the file uses the axis attribute for the offsets. I think ScippNexus already supports this, so we may just need to fix Chexus.

Did I understand correctly that the conclusion was that if x{yz}_pixel_offset has an axis attribute then it is fine that the shape is different from detector_number and chexus should not return a violation?

I'm asking because the file that failed the test /ess/data/coda/2025/994160/raw/994160_00040831.hdf does not have an axis attribute on the /entry/instrument/detector_panel_0/x_pixel_offset dataset, so the test will still fail if we fix this in chexus.

Did I understand correctly that the conclusion was that if x{yz}_pixel_offset has an axis attribute then it is fine that the shape is different from detector_number and chexus should not return a violation?

Correct!