That's a good point. I searched Perl 6 PEG and found it really powerful. Glad that PEG is now a core part of the language - it's rather exciting to write embedded DSLs with PEG, and if the implementation uses memoization, we could expect a better performance than native parser combinators ( and no worse than memoized combinators ) and a little more convenience than Frost's memoized combinators, which need some wrappers for handle real-world tokens.
I'm not sure whether PEG in Perl 6 can handle ambiguous grammars or not. Parser combinators were designed to parse natural language sentences at first but were also found useful for parsing programming languages later. It seems not so necessary for PEG to choose the first match if the implementation can endure the memory cost when generating the intermediate representation.