It's a bit like the old argument about seatbelts. People who didn't like seatbelt laws would identify particular accidents where someone who wasn't belted was thrown free of a car and survived, while the belted occupants died. And, once in a long while, those anecdotal cases may even have really happened. There might well have been people who'd have survived if only they hadn't been wearing their seatbelts. But those exceedingly rare instances were far outweighed by the unbelted people who died but would have survived if they'd been wearing seatbelts.
Or it's like COVID vaccines. There really are a few cases of people who died from nasty allergic reactions to COVID vaccines, or contaminated vaccines, who would have survived if they hadn't been vaccinated. But those exceedingly rare instances are far outweighed by the unvaccinated people who died but would have survived if they'd been vaccinated.
That's why, with all these things, it's important to understand the big picture from a data perspective. If we fixate on a tiny number of contrary cases, and set policy in a way that hurts far more people, that's going to be a tragedy. I'm not taking any position on where the policy should ultimately sit, but I'm saying we need to make that decision based on an acceptance that any policy will fail to work out well for some individuals, and so we need to be thinking in terms of picking among imperfect options to get the one that the evidence tells us will minimize suffering and loss.