Counts become incredibly slow on some databases like Postgres. It needs to do a sequential scan, which results in slow(er) performance on even a ten- thousands of rows. When operating at scale, this is very, very slow.
From what I understand it's not slow if it can use an index, but the criteria for that are rather hard to understand as a normal user. If you vacuum often enough and the visibility map is current, it doesn't need a full table scan.
