@Boshomi: How to see that problem / check that it still exists? (Also see https://mediawiki.org/wiki/How_to_report_a_bug )

@Aklapper Yes the problem still exists. supercount shows a delay of 21954 seconds, but supercount is not the case, there are other tools for editing like https://tools.wmflabs.org/giftbot/weblinksuche.fcgi which should work with very very small delay.

This is a known issue, 2 databases have crashed in the last 2 days. I have yet to investigate the core cause but I am taking care of fixing the replication issues first.

Please note that the fact that there are some issues with some replication channels shouldn't impact directly labs users, as not all are exposed as the default hosts (e.g. you should not use s1 on c2). See:
https://tools.wmflabs.org/replag/
For the lag of the ones exposed directly.

To give an advance of the work done:

Things are now getting corrected, and all replication channels are active again, but it will take some hours to reduce the replication lag back to 0. However, until the root cause is eliminated and due to some corruption in users's tables using non-transactional engines, we may have some hiccups in the following days.

A first evaluation points to some large imports going on a few hours before the crashes, but I cannot say for sure that is the real cause. I will continue investigating and reporting back.

https://tools.wmflabs.org/replag/ shows high replag again:

Shard Lag (seconds) Lag (time)
s1 0 00:00:00
s2 0 00:00:00
s3 122 00:02:02
s4 0 00:00:00
s5 17701 04:55:01
s6 0 00:00:00
s7 0 00:00:00
Wiki Shard Lag (seconds) Lag (time)
dewiki s5 17701 04:55:01
wikidatawiki s5 17701 04:55:01
aawiki s3 122 00:02:02
aawikibooks s3 122 00:02:02
aawiktionary s3 122 00:02:02
...

This is intended- see https://phabricator.wikimedia.org/T124172#1948046