Worst-case scenarios should be constrained by your knowledge. I know what's in the Pfizer vaccine, so my worst-case scenarios have to be constrained by that. But sure:
“Worst-case”, somebody's poisoning the vaccines. This leads to severe side-effects from some batches, which the statisticians take entire weeks to notice because they're all asleep on the job. Vaccine roll-out stops pretty-much everywhere, but nobody's able to trace down the saboteur. A dozen new, highly-transmissible and virulent variants emerge, with different antigens to each other and to the vaccines, so vaccinating the populace becomes intractable.
You see that this scenario is incompatible with “everyone gets vaccinated”. The bad stuff largely comes from “people don't get vaccinated”; the only way everyone*'s going to get vaccinated is if the vaccines don't suddenly get poisoned or otherwise become dangerous.
* except the immunocompromised, and others who don't get vaccines for health reasons.
I can't do a “worst-case scenario” where the vaccines are already poisoned, because they're not. It'd be bloody obvious if they were, given how big a sample size the statisticians have to work with. That would be even more blatantly fictional than my “tens of millions die every year” scenario.