@bojanlandekic@mastodon.social short answer is: snowflake IDs have a time-component to them, making them friendly for sharding and indexing. If you compress that into an alphanumeric string, you most likely loose the efficiency benefits. It's kinda like why you might see UUID primary keys not being recommended, when the standard UUID was UUIDv4 which had poor index & sharding efficiency, but with the newer UUIDv7, which are time sortable, you gain back that efficiency.
Basically randomness is the enemy of database performance, having stuff ordered by time in the primary key makes it easy for the database to know which pages of the database to look at, and which to ignore, as page files are created over time.