First forgive me if my termination is wrong, as I did not learn statistics in english
Assuming you cast a lot of shadowbolts (will refer to what's "a lot" at the end), you can say that your # of crits has a normal deviation.
Summing up a lot of shadowbolts, each shadowbolt has C chance to crit. Call X=1 if it crit and X=0 if it did not, and E(X)=C. Var(X)=E(X^2)-E(X)^2=C-C^2.
Casting N shadowbolts, E(average crit over N bolts)=E(X)=C (obviously, on average with 20% crit rate casting 100 shadowbolts you'd crit 20).
Var(average crit over N bolts) = Var(X)/N = (C-C^2)/N
The deviation would be Var^0.5.
For 100 shdowbolts and 25% crit rate, [(0.25-0.25^2)/100]^0.5=~5.6%, which means you have ~63% chance to have your crit rate between 19.4% and 30.6% if you only cast 100 shadowbolts. So getting 30% crit or 20% crit if you only casted 100 shadowbolts is very possible.
To reduce the deviation to 1% with 25% crit rate you need to resolve:
[0.1875/N]^0.5=0.01
0.1875/N=0.0001
0.1875=0.0001N
N=1875
So if you cast less than 1875 shadowbolts expect to get at least 1% deviation in your crit chance with 25% crit. Not to mention even with 1875 shadowbolts you only have ~63% chance to be within 24% to 26% crit chance and still have some chance to be a bit further than that, however you have well over 99% chance to be within 22% to 28% if you cast 1875 shadowbolts.
In other words, I wouldn't call "bug" unless your results are further than 3*deviation than the average.
As for the "many shadowbolts", it's because everything here is an approximation that gets better and better the more shadowbolts you cast, however it would be rather close to the accurate deviations if you cast like 80+ shadowbolts with a realistic crit chance. Cast more and it gets more and more accurate for calculating the deviation.