It happens in scientific computing with zero bounded signals fairly regularly. Sometimes values hover close enough to zero that the computation results in a/0.

It happens in scientific computing with zero bounded signals fairly regularly. Sometimes values hover close enough to zero that the computation results in a/0. It's not desired but the only alternative is a guard that invariably adds a performance penalty. There are other ways around it, such as normalization, but they're not trivial depending on the problem and can actually add other sources of bugs/errors. I understand the author's intent even though I disagree with the result.