Isn't the lesson of 2020 to just ignore the ACS and assume that if it's dramatically contradicting the census it's probably gonna prove wrong in the next census?
This feels like a generalization from one example. Prior to 2020, the ACS tended to be accurate, and
the Census has admitted that some of its counts in 2020 were flawed, with the ACS having been more accurate; the usual pattern was overcounts in Northern urban areas and undercounts in Sun Belt suburbs, though there are exceptions, which I think derive from how much particular states chose to cooperate, or not, with the Census.
2020 was a one-off failure caused by a mixture of the unexpected advent of COVID and Trumpian mismanagement, but I wouldn't really expect the same thing to happen again (and even if the ACS has errors in the 2020s, I wouldn't expect the error patterns to be the same as the ones from the 2010s, since the 2020 ones have been identified and at least on paper accounted for).