[RFC/patch] Fixing libstdc++/11378 and, while at it... ;)

[Benjamin Kosnik]

Eh. 

I don't really like this patch. I'd prefer not to see special casing for
unbuffered creep in again. And the bits that are there should be kept to
a minimum. The problem is real, however, so at some point some kind of
solution will have to go in: perhaps you could wait a bit on this one.

Paolo, you mentioned you have other plans for this, that impact all (not
just unbuffered sputn) calls. Perhaps you could do that first?

best,
benjamin