Hi all! weird situation - I have created new test dataset. running the test on this data create non-deterministic result for unmapped reads. so every --update-snapshot generate new md5sum. Have anyone saw such case? I wonder if this is specifically for the given tool or could it be related to the new dataset ?

Indeterministic output is usually the fault of the tool. You should check the mapper's documentation to know what the issue might be. As for the test, you should check the strictest thing you can while maintaining a stable snapshot – see module testing docs for details.