Ambisonics requires a lot of actual speakers to have a large (say, couch-sized) "sweet spot" where the directional audio sounds any good.

Really? I've heard some pretty amazing demos of ambisonics even with just two speakers - things which completely blow away anything I hear through mainstream productions with 5.1 and 7.1

But don't you have to hold your head in a smallish area for all the spherical harmonics to add up properly?

I find this curious, as you should in theory be able to map your 30 speakers to SH coeffs if you are using high-enough order SH (although 4th degree SH needs 25 coefficients so I guess there's not enough savings)

Like I said, that requires a lot of physical speakers, not just encoded data.