Some 18c woes

Shortly after upgrading to 18.5 some odd latch activity started to show up again for some of our modules:

SQL> @dashtop event "sql_id = '8ksqh62dm38ud'" "to_date('2020-02-23 02:15','YYYY-MM-DD hh24:mi')" "to_date('2020-02-23 02:30','YYYY-MM-DD hh24:mi')"

Total
Seconds AAS %This EVENT FIRST_SEEN LAST_SEEN
--------- ------- ------ --------------------------------------------- ------------------- -------------------
6840 7.6 39% latch: ges resource hash list 2020-02-23 02:15:03 2020-02-23 02:29:04
5410 6.0 31% library cache: mutex X 2020-02-23 02:15:02 2020-02-23 02:29:04
5350 5.9 30% 2020-02-23 02:15:02 2020-02-23 02:29:04
50 .1 0% gc cr multi block request 2020-02-23 02:15:02 2020-02-23 02:25:08
10 .0 0% wait list latch free 2020-02-23 02:25:08 2020-02-23 02:25:08

We ran into this previously in TEST and we'd thought it was solved it by setting:

_lm_res_hash_bucket = 32768

See Bug 29780459 : INCREASE _LM_RES_HASH_BUCKET AND BACK OUT CHANGES FROM THE BUG 29416368 FIX

This parameter had its default lowered from 64K for cloud and PDB customers to reduce the number of ges resource hash buckets.

We increased _lm_res_hash_bucket back to the default of 65536

UPDATE:

This has reared it's head again even with the 64K setting. Still in the process of troubleshooting, will update when I have a solution.