Brief explanation of the logic behind the changes:

ActionScript:
The lookahead was there to check if at least one of the 2 keywords (intrinsic or extends) was present. By splitting the expression into 2 expressions seperated by |, we avoid the need for this check.

BrighterScript:
Lookaheads where pretty much unneeded here.

FreeMarker:
The lookahead checked that the next character was either \s or ], but then it checks that the next character is anything but ], so the only thing that could have matched was a \s anyways.

RUNOFF:
Because the lookahead is right before a |, we can easily replace it by matching the content directly.

Solidity:
We can avoid the negative lookahead by using a character class for the next character that excludes numbers.

Hosts File:
Because the lookahead is at the end, we can easily replace it by matching the content directly.

@Alhadis, I think I'll let you handle the Vim Help File case, I'm not ready to deal with with nested lookaheads at the moment 😅.

@DecimalTurn The heuristic used by Vim Help Text was lifted from Linguist's Modeline strategy, which presents a much more readable and self-documenting version of the regex:

Any changes to Linguist::Strategy::Modeline::VIM_MODELINE will logically need to be reflected in the heuristic, and I strongly advise against making the regex even less readable by replacing nested lookaheads with… whatever works with re2.

In other words, you'll probably have to make an exception upstream for this particular heuristic. 🤷‍♂️

Ok, I'm going to take your advice on this one and leave it be for now.