That's fascinating, thank you! It looks like through blind luck I've been frozen-in-time with a version whose performance happened to complement the way I was trying to tackle the parsing problem...
For my side, I suppose I can stick with 5.8.8, or jump to 5.30+ by possibly changing my regexes as sw1 and SBECK have suggested to avoid "\s* ^ \s*" and compromising a little on the strictness of the parser.
But for the sake of Perl, however, is there a recommended escalation path to alert the developers about this regression / add it to the test suite / etc? Or is it niche enough not to be interesting?