It does scale badly in the bigger game modes which are played a lot more than 1vs1. The four main reasons are:
1. It has no real late game commander abilities (although cluster mines and vetted Paras still work well)
2. Greyhound comes way too late because of ressource inflation. At 3vs3 and 4vs4 it is completely useless, at 2vs2 hard to use already. It has to be integrated into tech and needs a small buff. As I said I would prefer a recon buff over a combat buff. That way it would have a greater team value too.
3. You want more Paras than howitzers. Reserve Paras should be buildable at HQ at least, a seperate drop would be even better, since it would open up a lot of combined actions at teamplay.
4. It has little team oriented commander abilities that have a unique value. Greyhound recon buff and seperate Reserve Fall airdrop would strenghten team options.
recon support scaling worse in 3v3 and up than other doctrines is certainly true, but at the same time, the inverse can be said about isu and elefant commanders that are absolutely meta in team games while rarely seen in 1v1s. i don't think this is much of a problem as not all commanders need to be viable in every gamemode.
that being said, i agree the greyhound could use some adjustments, especially with respect to its timing. your solution to tie it to tech sounds reasonable, though i'd argue a greyhound potentially arriving even before a stuart might be a bit too much. maybe 4 cp plus tech could work well, instead.
a buff to its utility via extended sight range is a great suggestion though. usf doesn't really have many great non-doc recon options anyway, so it would be nice if the greyhound could fill that gap at least to some degree.
with respect to the combat group... i don't know. the drop kind of acts as a soft cap that gives you access to both units at a hefty discount, while it dicourages spamming more than one or two at the same time. i do like this aspect of unit bundles in general (and this one in particular, since it is basically the only one where both parts of the package are equally valuable in almost every possible scenario), but i understand that i'm kind of in the minority here. buildable paras once the para drop is available would hence be a fair compromise i guess.