The solution I outlined is the one GitLab and PlanetScale both use internally, so it has been tested at scale and works well, for both Postgres (the former) and

The solution I outlined is the one GitLab and PlanetScale both use internally, so it has been tested at scale and works well, for both Postgres (the former) and MySQL (the latter). > It shouldn't contain information about the date, it shouldn't be auto increment, it should really be just random. That’s a great way to tank performance. You want your PK to be k-sortable.