It probably wouldn't come with any surprise that I favor Harry Partch over Cage as far as modern American iconoclast composers go.. :)
>It feels, though, that the practitioners are prone to getting caught up in the pseudo-novelty (novel, at least, to them) of how it is done at the expense of creating music that presents some engaging development arc.
Indeed, this is the crux of the matter / critique as that was very apparent from the get go of the live coding (LC) movement. I think LC is a tool which can be useful at times for exploring different sounds, but is not well suited as a performance mechanism. There is no or little apparent composition process involved in what I have seen over 10 years of LC. The mechanism in itself does not make lasting art, but if used at bare minimum via the process of selection as part of a larger compositional process it is valid. I'm sure somewhere in the LC community this is happening, but it certainly isn't how the status quo operates or how LC is presented at large.
In regard to TOPLAP et al; if you need to lead a "movement" with a manifesto to justify the musical output you're doing it wrong... Or in other words this movement like many before with leanings toward the avant-garde does not address that at the end of the day it's a handful of folks performing for other LC aficionados and no one else. One shouldn't need to buy into the manifesto to appreciate LC as art.
In many ways by elevating LC to "movement status" it can capture and beguile a generation of computer musicians including HN readers who are new to computer music concepts & generative / algorithmic music. It's not the only game in town so to speak though I do have general respect for the immediatism aspect of LC as far as it aligns with the mechanism and performance in regard to just getting out there and "doing it".
My experience and lifes work on the coding / performance side could be considered an antithesis to the LC movement; it's been a solo journey for me so far. My work has not been released yet though I may have been one of the first to build an alternative client / framework for SuperCollider ('02). It all started there and work continues today though it's branched beyond solely music / audio control, but I do hope to get back to music / performance applications and heck performing once the larger effort is released (and shock / gasp perhaps I'll make income to support the art side of things and future music-tech efforts).