To be concrete, if I use PG 10.x master-slave replication in production,
is it very reliable when set up properly?
It is more reliable than default MySQL replication. Default MySQL replication is logical whereas default postgres replication is binary. In the crudest of terms, if something goes wrong you can't just 'skip' statements in postgres to get replication to work like you can in MySQL. This ensures better data integrity during replication. As for non-defaults, pg10 and onwards support logical replication as well, with the usual caveats that logical replication comes with in any database.
I am a happy PostgreSQL user, but I have never tried PG replication at all.
Based on my vague recollections concerning the past, MySQL used to have the advantage of having a pretty stable and working master-slave replication when PG did not have such a thing. But PG has been more SQL-compliant, has had foreign keys forever, and in general, I guess PG has had a more advanced feature set available.
I have followed the progress of PostgreSQL regarding replication just occasionally, but not very actively. Master-slave replication is now officially supported in the newest releases, maybe has been for a long time, I am not sure.
In any case it is not my intention to spark a heated discussion about the merits of MySQL/MariaDB vs PostgreSQL. All I am asking is maybe hearing some observations of PG admins out there.
To be concrete, if I use PG 10.x master-slave replication in production, is it very reliable when set up properly? Any practical experiences? Well, I suppose it *must* be reliable, but I am interested in hearing of possible problems, too.