And another opportunity to simplify our code between different
PSGI-ish implementations. The snapshot retrieval is simpler,
but potentially slower since we waste cycles scanning for tags
even after we've found one. It's probably not a big deal since
it's only short info lines and we can utilize pipelining.