Literal survivorship bias.

There may also be a hard-to-infect/hard-to-seroconvert portion of the population left that could practically walk unmasked through a COVID ward in a hospital and not get infected. If they exist then the seronegative proportion of the population will start to be composed mostly of them over time through "purifying selection".

This is also why I'm somewhat skeptical about all the vaccine efficacy numbers that studies are throwing around. I don't think an adequate control group exists any more.

That's a good point: having a control group is pretty much a must to be able to get reliable numbers on efficacy and any future vaccine or medicine is going to have to some extent work with corrupted data from day #1.