Daniel Vérité checks the version:
With three locale providers (libc, icu and builtin), a PostgreSQL instance has potentially three different versions of Unicode at the same time.
When doing only store/retrieve or basic operations on UTF-8 strings, these versions don’t matter. Postgres checks that the bytes in the strings constitute valid UTF-8 sequences, but not whether the code points corresponding to these sequences are assigned or not.
It starts to matter when interpreting these code points as characters.
Read on to see which features might determine which Unicode provider you might be using.