Sequence numbers are a promise. Read the damned promise.

You have a bigger sequence number[1]Ref. 1sequence number rfc-editor.org than you had before, and you're ready to delete the earlier file. I'd hang on to it a little longer.

I want to know who issued the number, what it counts, and when it can start over. Without those answers, comparing two numbers gives me very little evidence that everything earlier arrived.

I would keep the missing range separately from the highest number observed. A single high-water mark cannot tell you both how far you've gotten and which holes remain. Let a newer file automatically close an older gap and you've declared success without learning a goddamn thing about the file you couldn't find.

Record counts need the same care. You can have a complete file sequence made of incomplete files, or a complete file with rejected records[2]Ref. 2records en.wikipedia.org. Sequence continuity proves one property of the transfer; there's still accounting to check. The numbers can line up beautifully while I'm still looking for the damned records.