My actual main solving method is already an amalgamation of like five methods so, uh, I guess????
Also: why are we calling the general form of method neutrality "option select" now? I thought OS was specifically for like, last layer or LSLL methods (or whatever finishes off a solve). Am I just out of the loop?
Option-select is a more specific form of method neutrality. It's sorta like how you do an easy WV or a ZBLL if it pops up during a CFOP solve except you actually use 2 variants of a method actively. You choose which variant to use depending on the situation. While OS is worth it if the difference between the variants isn't big, Method neutrality is not worth it at all(Believe me, I've tried it)
(note: I rewrote this reply; it was originally needlessly sarcastic)
The way I see it, method neutrality requires you to be good at multiple methods. What even constitutes a
different method is, obviously, a fuzzy line, and where you draw that line could well decide how much of method neutrality is worth it for you.
At one extreme, if you consider CFOP with different cross colours to be different methods (I don't think anyone seriously holds this position, but it's always good to push the arguments to the limits to see where they break), then colour neutrality can be thought of as a weak form of method neutrality, and most people consider CN to be worth it. The "different" methods here are still similar enough that time spent improving on one method will help with another.
At the other extreme, you could have completely unrelated methods (say, CFOP and 3-style). The dissimilarity of the methods means that it's considerably less likely for a scramble to simultaneously be bad with both methods, but it also means that there's a lot less learning transfer between the methods, reducing improvement efficiency. In the short run, it would look like this type of method neutrality isn't worth it, but over much longer time scales (years/decades of practice) I think it's much less clear – if you've plateaued with one method, method neutrality lets you supplement that method's flaws with a different method.
And then in between you have all the half-hearted compromises like switching between different last layer methods depending on case (OLL-PLL versus OLLCP-EPLL versus 1LLL subsets etc.), and the like. The amount of learning transfer is in between the two extremes, and also the usefulness of neutrality is
also between the two extremes.
