The reason you cannot split test a smart sleeknote is actually quite simple. In some ways the smart triggering works like a split test in that it uses different triggers to find out what gives the best conversion rates for that specific sleeknote.
This means that there is actually a range of different sleeknotes (with different triggers) within a single sleeknote.
Apart from the technichal difficulties this present, it also means that a A/B split test of a sleeknote with smart triggers enabled will not yield any meaningful results.
A quick example:
Sleeknote A has smart triggers enabled
Sleeknote B doesn't have smart triggers enabled
Sleeknote A with trigger 1, has a conversion rate of 2,1%
Sleeknote A with trigger 2, has a conversion rate of 3,6%
Sleeknote A with trigger 3, has a conversion rate of 2,9%
Sleeknote A with trigger 4, has a conversion rate of 1,8%
Sleeknote B, has a conversion rate of 2,7%
As you can see this means that Sleeknote B is performing better than Sleeknote A with triggers 2 and 3 but worse than Sleeknote A with triggers 1 and 4.