The equation is something like this:
If time-spent-adding-programmer > work-new-programmer-produces then
Recently occurred to me there is a corollary to the idea I hadn't previously considered:
If time-spent-code-reviewing-programmer > time-another-programmer-would have-fixed-issue then
In plain English, if the time spent by a programmer producing buggy code plus time spent by other programmers in code reviews is greater than the time other team members could fix said bugs, your project ships later.
It seems like the right corrective action is to fire the underperforming developer for the project to not fall further behind.
Shorter version: keeping a bad programmer on a project always makes it later.