Some mistakes end up being - ho-hum kinds of things, and the situation can be fixed - either a temporary workaround or a permanent correction. But others may be worse, might not be easily patched up, and may make a difference in the election outcome - an intolerable result. Worse yet, they might make changes in the outcome that do not get caught - or not caught until "too late".
Here is a current, real case, of such a computer program mistake, in my own county.
We use vote-by-mail. Ballots are printed and mailed to voters. Voters mark them and mail the ballot back (or drop it in a designated collection-box.) The county is geographically divided two different ways - by precincts and by special taxing-districts, like school, fire, cemetery and parks districts. Their boundaries don't coincide, so many precincts have sub-sections commonly called "splits". Voters in one split can vote on some offices and measures, voters in other splits, cannot.
This year, one of the companies changed their program. The company who made it either didn't know about, or didn't adequately check, how it interacted with another program. The result - a failure. And yet the changed program (we are told) was passed - approved and certified by the Federal certification process and the State certification process. Why was the error not caught there? Duhhhh!
So, as our county was preparing the ballot-images for our OpScan ballots to be sent to the printer, two of the programs goofed in interacting, and the computer output had errors. The program which was changed had a new identification way of registering the precinct-split number, but the other program that received that data did not know the new field identification.
Only when the next step was being done, setting up the ballot-reading program, did a third program squawk - because it saw that those ID's were missing. Aw, shucks. It took the staff several days to trace the problem, and to craft a solution.
Maybe that will get fixed well enough for this election, already running. And surely well enough for the next election - until the next mistake appears.
This is the main reason I do not have a high degree of confidence in computerized voting and vote-counting. The potential - nay, certainty - of errors is too high to engender the degree of confidence needed, thus the trust that the public must have, in the use of elections as the way to let the people control the government.
The opposite is intolerable, and the founders of our nation knew that, and wrote it into the Constitution of the United States.
Maybe we can continue to use computers in our elections systems, but ONLY with a parallel auditing program that can be independently verified totally separate from the basic system. And since I reject the use of yet other computers to do this, the only such process which I know to check the accuracy and integrity of the vote-counting is by hand and eye. This is why I strongly support the standard use, everywhere and always, of a program of hand-counted double-checking of the ballots and vote-counting.