Nanoid’s performance is likely to be on-par with UUIDv4, if not worse. While it’s smaller in ASCII, Postgres and MySQL (when coerced) both store UUIDs as 16 binary characters, well below the 21 of Nanoid.

And in any case, the primary driver of the poor performance is from k-sortability due to page spread. The only alternative to UUIDv? that would do well is ULID, as it’s lexicographically sortable.