This goes back many years, but among the many protease inhibitors that have (mostly successfully) treated AIDS we had the following mutations:
D30N: Nelfinavir. (Agouron/Pfizer).
M46I/I47V/I50V: Amprenavir (BMS).
L10R/M46I/L63P/V82T/I84V: Indinavir (Merck)
M46I/L63P/A71V/V82F/I84V: Ritinovir (Abbott).
Saquinavir: G48V/L90M (Roche)
cf: Antiviral Drug Discovery Summitt, March 28-29, 2001
Protein Science (2000) 9: 1898-1904
In HIV, there is no transcription correction mechanism, and there are about 10 billion viral particles produced in a day. Of course many of the mutants are not viable, but those that are survive and reproduce more readily since there is selection pressure given that the susceptible viral particles do not reproduce.
It is unsurprising that in active infections, antibodies show the same effects. The point is that the infection must be active for this to happen. If the antibodies, as with a vaccine, are effective, there aren't 10 billion particles (or however many SAR-CoV-2 produces) and thus a lower probability of errors responding to selection pressure.