to speak for humble me,
a) i am constructing one diachronic a posteriori language (nassian),
where the protolang is somehow constructed too, but is real (early
common slavic)

b) and one diachronic a priori language (arkian), where the
historicity is fully constructed, even the protolang is constructed

within both, i have started with some preliminary ideas what the sound
of contemporary version would be. ie. in nassian i wanted no [j], no
[v], no palatals, lots of umlauts and geminates. then i had to find a
logical way to it, respecting natural sound-speech changes. as you
know, they were many re-iterations of it as few of my intentions were
unreal and few appeared to be necessary although not wanted. but the
major problem for me was to reconstruct the protolang, cos early
common slavic is not hobby of majority of slavists and me is poor/pure
amateur. to create a historical phonology was relatively easy. more
complicate - historical grammar. the worse was and still my nightmare
- historical lexicology, ie. line of succession of meanings and
related words (ie. in real slavic languages the original word for
"member of our tribe" became word for "foreigner"; there are tons of
such meaning shifts and they make language real). hell difficult and
almost impossible-to-crack problem.

arkian was easier. there was the protolang, Hiberian, constructed by
my friend and it was a solid constant in my work. i only have to find
a way from hiberan to arkian, which is even more easier if i assume
that the time between them is short and arkian is constructed language
per se, in the definition of its construction scheme. nassian has vast
time period between protolang and actual state (1500 years), is more
"real", should be logical, quasi-natural and believable.

i am thinking about to write bachelor thesis in common linguistics on
construction of fictive diachronic language, cos it is appealing problem.

hope it helps or is interesting anyway.