What the county sees and what the state sees can be different especially when obscured by summation.
A county can check their numbers and be satisfied. But upon transmission to the state, the DISTRIBUTION of numbers from many counties within a sum can be changed by weighting, even as TOTALS still add up.
103 = 50+53 = 53+50 = 103
A checksum function doesn’t catch this.
When a county looks at the number they sent, they see it is, in fact, the number they sent. But at the state level, the software weights the number to something else while PRESERVING THE SUM TOTAL of all counties together.
It’s a clever trick and it can be remedied but an EFFORT NEEDS TO START NOW.
The only sum being preserved in your example is the total. And I'm not referring to only a repeat of the simple addition.
County A reports 50 votes for Candidate X, and 53 for candidate Y.
State uses 53 votes for candidate X, and 50 for candidate Y. Nothing subtle about that. Candidate X is focused on HIS vote count, and County A ought to detect this transposition (I know it need not appear as transposition error) without any help.
I have seen the voting machine trick, where the starting count of 0 is obtained by preloading +100 to the destined winner, and -100 to the destined loser, but that bias is hidden from view. Vote subtotals though, starting at the precinct level, are visible.