Good general rule. However, if you are doing some kind of general economic modeling, with estimated parameters all around, not claiming to represent anyone's accounts or ledgers, I suspect an exception applies.

"When you are dealing with things that are measured" ... or wild-ass guesses.

And I doubt demographers and actuaries are all using integer arithmetic for projecting technically integer-count human-life outcomes.

Calculating taxes: Be careful and conventional, as you say.