I think if your system for finding the top 8 playoff teams is to identify the top 8 "best" teams, and you choose W/L as your primary metric for determining team skill, using RW% as a secondary metric is not insane. After all, if your goal is to just identify the best 8 teams, why do I care about how you did vs. a team that wouldn't even be in playoffs by this ranking system. I would care more about how you did vs. the division as a whole. This is especially true if you use the (what I think is) ideal definition of team skill, that being the the average probability of defeating any other team in the division.
That being said though I still think H2H is the way to go for tiebreaks because the forced map schedule makes the variance during the regular season way higher than if it were pick/ban, even with 16 matches compared to something like highlander's 7/8. Makes me more inclined to say fuck it and just reward W's over everything.