While only >10%-share browsers can be worth commercially developing for

That's one problem: many web developers program for browsers, not standards.

    In other words, ideally, there would be a "polyculture" in innovation and experimentation,
    that would constantly funnel into a "monoculture" in actual final standards.

Yes, and IMHO that is what happens to browser rendering engines atm. Everyone agrees/complies to one technology and by doing so it becomes a de facto standard. Then new approaches for new problems can be taken from there, without reinventing the wheel. But this also creates room for alternatives to rise, so we are back to multiculturism. :)

E.g. replacing JS with some other language would be such a new alternative. Open-Source-Lua-Presto, imagine that! ;)