In the discussion with @daniel and @Anomie it become apparent that it's better to drop support for Oracle and MSSQL before trying to abstract it and moving support to extensions and if someone wants to maintain the support, they should do it in the extension.

By the way, I have another bug to report. This one I have not yet traced in the source (it is very hard to get it to actually happen and thus get a stack trace): unquoted column identifier for _pageID, once again.

Change 512380 merged by jenkins-bot:
[mediawiki/core@REL1_33] resourceloader: Use AND instead of OR for upsert conds in saveFileDependencies()

Change 512396 merged by jenkins-bot:
[mediawiki/core@REL1_31] resourceloader: Use AND instead of OR for upsert conds in saveFileDependencies()

Change 512395 merged by jenkins-bot:
[mediawiki/core@REL1_32] resourceloader: Use AND instead of OR for upsert conds in saveFileDependencies()

Change 512397 merged by Reedy:
[mediawiki/core@REL1_30] resourceloader: Use AND instead of OR for upsert conds in saveFileDependencies()

Change 512397 had a related patch set uploaded (by Reedy; owner: Reedy):
[mediawiki/core@REL1_30] resourceloader: Use AND instead of OR for upsert conds in saveFileDependencies()

Change 512396 had a related patch set uploaded (by Reedy; owner: Reedy):
[mediawiki/core@REL1_31] resourceloader: Use AND instead of OR for upsert conds in saveFileDependencies()

Change 512395 had a related patch set uploaded (by Reedy; owner: Reedy):
[mediawiki/core@REL1_32] resourceloader: Use AND instead of OR for upsert conds in saveFileDependencies()

Change 512380 had a related patch set uploaded (by Reedy; owner: Reedy):
[mediawiki/core@REL1_33] resourceloader: Use AND instead of OR for upsert conds in saveFileDependencies()

Change 512301 merged by jenkins-bot:
[mediawiki/core@master] resourceloader: Use AND instead of OR for upsert conds in saveFileDependencies()

Thanks for the report @IijimaYun. I believe this bug was caused by e7b57d881a, first released in MW 1.29.

[11:17:22] <Remilia> Reedy: can confirm the patch totally fixes the deadlock issue, zero deadlock errors since I added it on the main wiki (11+ hours now) while the other, not patched yet, wikis get them (once a day since they are very low use)

Change 512301 had a related patch set uploaded (by Reedy; owner: Reedy):
[mediawiki/core@master] Make md_module, md_skin a single unique index in ResourceLoaderModule::saveFileDependencies()

MySQL is doing (using dummy values)...

INSERT INTO `mw_module_deps` (md_module,md_skin,md_deps) VALUES ('skin name','vary','deps') ON DUPLICATE KEY UPDATE md_deps = 'deps'

Update: this seems to trigger DB server overload. Deadlock detection has timeouts and this leads to PostgreSQL reaching its connection limit (my max_connections is set to 1024 and it gets exhausted during massive user influx).
These hanging connections seem to lead to ResourceLoader actually failing to load CSS and JavaScript.

@IijimaYun - I'm sorry for the very long delay, and I'm sorry about all the problems. I assume that issue you had with "UNIQUE" is due to the fact that you have some template ("ShipMain", I guess) that declares more than one Cargo table. Which is fine - but it's rare, which is probably why this problem hasn't shown up before. I checked in to Cargo a fix to remove that "UNIQUE" restriction, as well as that strange "DEFAULT nextval()" thing. If you can do the same thing locally in your database, it should be fine. (You probably already have.)

@siebrand , this UserMerge issue is specific to PostgreSQL; do you advice to migrate to MariaDB or wait for a fix?
I hope you can help.

Thanks @Reedy , upgrade was from 1.27.1 to 1.32.1

Fixing part of this would also fix T221616: Error running update.php on shared db with PostgreSQL (at least where INSERT IGNORE is concerned)

Hacking the latter syntax into maintenance/populateExternallinksIndex60.php like so
allows the upgrade process to finish successfully.

OSM is PostreSQL which is not maintained by the DBA team.

Patch above + its dependencies and that's it.

Change 503562 had a related patch set uploaded (by Daimona Eaytoy; owner: Daimona Eaytoy):
[mediawiki/extensions/AbuseFilter@master] Make patterns searching compatible with Postgres and SQLite

Calling this resolved because the compatibility problem with postgres is indeed resolved. I'm gonna open another task for the long-term goal of splitting afl_filter in afl_filter_id and afl_global.

Change 468706 merged by jenkins-bot:
[mediawiki/extensions/AbuseFilter@master] Use string cast for Postgres compatibility

In T191231#5067674, @daniel wrote:

@Ladsgroup Did you check how DBAL holds up against the heap of nasty edge cases that @Anomie compiled? https://www.mediawiki.org/wiki/User:Anomie/Abstract_schema/DB_Requirements

Change 488477 abandoned by Daimona Eaytoy:
Read afl_filter_id and afl_global instead of afl_filter

@Ladsgroup Did you check how DBAL holds up against the heap of nasty edge cases that @Anomie compiled? https://www.mediawiki.org/wiki/User:Anomie/Abstract_schema/DB_Requirements

Change 488477 had a related patch set uploaded (by Daimona Eaytoy; owner: Daimona Eaytoy):
[mediawiki/extensions/AbuseFilter@master] Read afl_filter_id and afl_global instead of afl_filter