>Anecdotally, I know someone who has been infected twice

How does he know he was infected twice?

Saying that a PCR test returned positive values two times isn't a great answer - since by its nature PCR is only counting the existence (above the programmed threshold post amplification).

Serology is the only way to know for certain you were/are infected.

This highlights one of the bigger issues we have: PCR test are a very cheap and a useful way for getting estimates but their failure rate is high enough to make it pretty poor basis for doing large scale data analysis.