Re: consensus call of RR prefix

[...] leaving the TXT records in the domain.tld dataspace, while ugly, seems like a good bit of parallelism for the SPF records that will be published in the same place.
I concur.  Purely pragmatically, many people will want to check both SPF and IETF-compliant Sender ID.  Being able to fetch both with a single query is therefore beneficial.

... but also impossible if the resource records are, as here, of two different types - TXT and SPF, respectively.  (Before someone says it: No, "any" is not the same as "all".)  In the general case, which applies here, to look up two resource record sets of different types always requires two DNS transactions.

(Of course it doesn't matter if I use separate pieces of software to check SPF and Sender ID, both of which make their own queries; my local caching nameserver will still ensure that only one query is made to the authoritative nameservers.)
  

That is not what will happen.  Front-end transactions to proxy DNS servers only combine at the back-end if they are looking up the same resource record set keyed by the same <owner,class,type> tuple.