UUIDs also introduce performance challenges. Using them as the primary key will absolutely torpedo your insert performance. They also subtly harm performance on

UUIDs also introduce performance challenges. Using them as the primary key will absolutely torpedo your insert performance. They also subtly harm performance on pretty much every other operation. You can fit half (or 1/4, if you can live with 32-bit ints) as many UUIDs in a cache line, and can't even fit one of them in an integer register. The impact on join performance can be pervasive at production scale even when it's not visible at dev scale. And UUIDs tend to result in records being physically arranged in ways that defeat the page cache.