PEG are not silly at all. PEG *include* regex, but not viceversa,
and it seems to have the same power of  CFG, so you can build a parser for pratically anything
(and you can not do this with regex) .
Perhaps the reverse sentence has more sense: drop the "silly" (?) regex, use the new peg.
For what I know the reverse side of lpeg is that a file must be loaded in memory completly.

Maybe columnset can do  more, or we need more examples from hans .

why not xquery too? Are mkiv  supposed to be a full  xml processor ?
A bit of xslt can be done with lpeg (and not regex),
but usually  it's better to prepare a xml to typeset before processing with mkiv
("use the right tool for the right task").

and worse too .
Things are now a lot better than before ,IMMO .