Why many more deadlocks after upgrade to PG 17.5?

RHEL 8.10
Prior version: 14.18

There were deadlocks when at PG 14, but a small fraction of the current
number of deadlocks.

All tables were vacuumed and analyzed immediately after the pg_upgrade. 😉

Tables are partitioned by range (weekly). Physical replication; no logical
replication.
There have been no code changes since the pg_upgrade (performed 9 nights
ago).

Attached is a section of the PG log file. It's the same kind of deadlock,
in the same code as before; just now there are _more_ of them.

I don't control the schema or the application, or the code in the
application; we just need to know why there would be _more_ in 17.5 than in
14.18.

--
Death to <Redacted>, and butter sauce.
Don't boil me, I'm still alive.
<Redacted> lobster!