Thanks for the review @SPPearce, I don't fully understand why the conda test is failing as the tool is on bioconda and deployed as a biocontainer. I would assume that's an issue. Should the conda test be skipped?

There is presumably something inside the test_all.ome.tif file that is different between docker/singularity and conda.
This happens often for some file types, such as bam files, where the header of the file encodes a file path that is different between the two environments (i.e. /opt/bin instead of /usr/bin or something).
Generally at that point we test for existence of the file, or the presence of a string rather than the entire file.

The issue I'm seeing is that the conda output matched the md5sum in the previous test, even though the environment definition was the same.

I'm also noticing an issue with application of flat and dark field correction - should be looked into before merging.

A short update, I've just tested and a newly built Seqera container with ASHLAR from bioconda produces the same outputs as the conda version, but I haven't yet found the precise reason for the difference.