Metric evidence is surprisingly hard to find.
According to my corpus searches, almost all examples of quoniam in
Vergil,
Horace, and
Catullus
are in hexameter and not in the second last foot.
I don't quite know why; I see no obstruction to having the word that late in a verse but it seems not to be found there.
In these contexts both readings scan well.
The only exception in these three poets is in Catullus 61 (lines 194–198):
non diu remoratus es,
iam venis. bona te Venus
iuverit, quoniam palam
quod cupis cupis, et bonum
non abscondis amorem.
The syllable length pattern is (glyconic & pherecratean):
x x — v v — v x
x x — v v — v x
x x — (v v —) v x
x x — v v — v x
x x — v v — x
Here —, v, and x stand for long, short, and either length.
I put the relevant three syllables in brackets as the bold-faced font is not very distinctive.
Conclusion:
In 23 of the 24 cases in the mentioned three poets quoniam can have two or three syllables so both j and i are possible.
In one case only the version with three syllables is valid and thus an i is forced.
Whether this is poetic licence or the typical pronunciation, I cannot tell from this one example.
But it does suggest that the vowel i is the general pronunciation.