MW schema has, for revision:
CREATE INDEX /*i*/page_user_timestamp ON /*_*/revision (rev_page,rev_user,rev_timestamp);
Several old wikis (enwiki included) do not yet have page_user_timestamp. After trialing it on enwiki slaves, it does not seem to offer much benefit over page_timestamp, and sometimes causes enough optimizer confusion that some normally fast queries jst time out. If specific queries do use it, perhaps those should be on a pager slave... or something.
This may be something that can be better controlled with another tilt at persistent engine-independent index statistics, but OTOH the chance to remove an index from an already-large table like revision is attractive. Need to investigate how the small/newer wikis are using the index (or not).