The preambula:
It appeared that tag remapping after we received response from t5memory works depending on the order of matches in response.
We send requests to a split TMs in parallel and responses can go in different order.
If segment target contains same tags as in segment source, but with different rid for some reason (e.g in source tags are 1,2,3 and in target 4,5,6) source tags should be normally reapplied to target by tag handler (t5memory_xliff)
And in most cases it works properly.
But if before the 100% match segment (which contains such tags with different rid) we have some segments with excess tags with same rids tags reapply mechanism starts working improperly and as a result we have the following situation:

• in segment source we have tags 1,2,3
• in t5memory segment target we have tags 4,5,6

Expected behavior should be that as a result we should have in segment target tags as in source 1,2,3
Actual behavior is that we have tags 4,5,6