You don't think ongoing performance issues and bug fixing are problems? If the MooTools selector engine were entirely bug free (I'm not claiming that it's full of bugs, just that without knowledge to the contrary I'm going to assume that it's like most other software projects and has at least one or two) and as fast as Sizzle, then yes, changing engines would be more work for no obvious benefits.
However, if they can get measurable performance improvements and increase the number of eyeballs on the bugs on that one portion of their code, then that seems like a win to me. There are plenty of other reasons why they still might not want to integrate Sizzle, but this one doesn't strike me as compelling.