First of all, this seems like an entirely pointless endeavor for the simple reason that, by mandate of the LLG, no change proposals are allowed until the baseline as is has finished being fully documented. Obviously there have been a very very few exceptions to this, most notably xorlo (, actually that's the only one I can think of other than (possibly) dotside), but I contend that any such changes are of a "bugfix" type of change. There are some proposals I am aware of that I have no doubt would be approved, such as changing to using PEG instead of YACC, .camxes. instead of jbofi'e, etc., but of utmost importance to any proposal, past, present, and future, is getting that baseline finished documented so the freeze is removed. With that freeze in effect, it really doesn't matter /what/ the proposal process is, or who the arbiter is.
Secondly, to the best of my knowledge, the current proposal process is to create and submit a specification on the proposal to the BPFK, who would then review it, assess how it would affect the corpus, affect any changes deemed necessary, and then vote on it for approval, while Robin holds the role of, basically, "Super Veto Man", in a very similar vein to the way Congress and the President operate RE: the passing of laws in the U.S. It sounds to me as if you're suggesting giving the role the BPFK plays in the process to selpa'i, not the role Robin plays, which I don't approve of. If, on the other hand, what you are suggesting is that selpa'i be the intermediary between the proposer and the approval committee, then I see nothing wrong with that. (And before anyone says anything about the BPFK being dead, may I just point out again that the role of the BPFK is that of maintaining the language- i.e., documenting the language as is as well as approving and recording any changes to it. Since there can not be changes with the freeze-until-baseline thing is over with, that kind of leaves the BPFK with nothing to do. All I can say on the subject is, there's a well-documented process for finishing the baseline, and anyone that wants to can easily go about helping to finish it, and it is about 9x% finished as of this writing, so it's not like there's a lot left to do.)