I have had this issue with vengeance multiball on my STLE since new. (code 1.3 then 1.4 & 1.41) With info from posts above I have tried several controlled experiments in a basement labratory setting and have conducted a long list of tests containing very tight tolerance limit testing, all being observed by my 9 and 11 yr old assistants. We have concluded that our machine will release the ball on the magnet with a light "pass" across the first opto. With either flipper in the up position the pass across the opto will not release the ball.
If the ball is hit hard enough to push the first ball up, the second ball would usually stick on the magnet even if the first ball was hit hard enough to bounce off the back of the vengeance alley then stick to the top of the second ball that is now on the magnet...
I have not installed the latest version, 1.42 of the software.