Random deadlock during Solr cores initialization

From what we know, one of the requirements to reproduce this is having several cores, yes.

But I don't think we ever noticed this problem when we had only 2 cores (search and events), it probably really started more with the introduction of a 3rd one (like in 12.7 which was then merged into a more generic rating in 12.9). AFAIR there was also huge differences of reproduction rate with the version of Java you were using (AFAIR it was not happening much with Java 8 and really started to be noticed when we started to use Java 11 more).