It’s of course possible that “test data” was an excuse they made up after the fact. Maybe it’s a cover story to hide even worse incompetence (or maliciousness). Why did it require a candidate to notice that the sums between rounds were wildly different?

And wtf is a test ballot anyway? Are these physical ballots, or entries in the database? Neither inspires confidence in the integrity of this election.

> Also, prod and staging/QA/dev - never the twain shall meet.

You're expecting a lot from developers who were chosen because of their relation to elected officials rather than their domain expertise.

> The list of relatives stretches even to the [Board of Elections] computer programmers, including Rubén Díaz III, a Democrat who is the son of the Bronx borough president and grandson of a City Council member.

https://www.nytimes.com/2020/10/26/nyregion/nyc-voting-elect...

The flipside of this is that some defects can get through because assumptions made by the data used for QA/development might not be true for production but you don't find that out until the code is actually in production working with production data.