I really like that idea.

There's just a minor point of bikeshedding: should it be (?s:.) instead of a simple dot?

I can argue both ways: negated character classes include the newline (unless it's explicitly listed as excluded), and it becomes more predictable that way.

On the other hand you gain more flexibility if respect the /s setting of the outer regex, allowing the caller to tweak its behaviour accordingly.

The first point sounds more convincing to me, but I have to think a bit more about it.