I am guessing
@rzaneti means concurrency not parallelism, aka go into settings and turn it on, but set it to 1, so only 1 can run at a time, otherwise you will get multi-threaded deletes happening at the same time.
I also always recommend that you delete rows in reverse, starting at the latest working your way backwards.
You said it works before, and while I do not doubt you there are many things that can make something different, without it being "different".
Another 2 things
1. I would add a Try / Catch error handling scope around all deletes, this way you can validate for sure if something did, BUT, it might once but probably not more than 1 time.
2. Looking at your flow Run, do you see any consistency, where its always row 167 that somehow has an issue.
The reason that I say go from back to front is, that way lines do not get re-saved as a different id that you haven't gotten to yet.
So going from
764
763
762, you never have to worry about deleting 762 and 763 becomes 762 etc.