Jumping in rather late here (apologies if this verges into thread necromancy) -- I just wanted to say I appreciated reading the earlier post here suggesting the 'success' or 'reward' of working up a conlang from a reconstructed (nat-)protolanguage lies in the ability to compare and contrast it to its actually documented relatives, both linguistically and 'historically' (in a broad sense). I definitely feel the same way!
Unfortunately -- despite an undergraduate degree in Sanskrit
-- I really have no taste for the IE family, nor background in European history/language history, which puts a good 80% or more of a posteriori conlanging discussions (let alone group projects) out of my league.
FWIW, in my own tinkering I think of there being two related but different approaches. One is to work from a (reconstructed/hypothetical) protolanguage of an actual group of natlangs, to create a new sibling -- in which case I'm relatively less concerned with being "rigorous". (Though only relatively). At least for the dismal, petty language families I favor, it doesn't take long to learn how tentative and flimsy the reconstructed protolanguages are*, and how perverse it is to limit yourself to 'proper comparative-historical linguistic rigor' in inventing a new descendant.
The other is to take an older, historically documented language and work up a more or less direct descendant of it. (E.g., a descendant of Old Georgian found in some village in Mazandaran, or a 10th century manuscript in a descendant of Gaulish). I feel like that calls for a lot more rigor, in having a good grasp of the source language on one hand and the later linguistic and social environment it survived through.
[* Assuming much reconstruction has even been done!]