In some of our (zhwiki) filters, I use something like:

(removed_lines + 'UNIQUE_STRING' + new_wikitext) rlike '<ref name="([^"]+)">.*UNIQUE_STRING.*<ref name="\1"/>'

Thanks; while I am trying to get that to work, I still want to point out that this method can only pass the matches to the next part of the same regex (i.e. \1 only works within the context of the same rlike function). Ideally, one should be able to start \1 into a variable and then use that variable in a subsequent line of code.

(In reply to Huji from comment #2)

Thanks; while I am trying to get that to work, I still want to point out
that this method can only pass the matches to the next part of the same
regex (i.e. \1 only works within the context of the same rlike function).
Ideally, one should be able to start \1 into a variable and then use that
variable in a subsequent line of code.

Yeah but normally this trick works everywhere where the variable is used a next regex match, because you can always concatenate other variables together and use a big regex to match that long string, like what's done above.