still scrutinizing the vid/looking for insight but i believe i'm seeing plenty of tunneling (was focusing on the RH flipper) while assuming copious use of SetSpritePhysicsIsBullet(().
as we don't have 100% (much less current) box2d implementation, consider your own prediction/time of impact (TOI) functions (to account for flipper rotation, especially).
i have some basic routines in place
here but since i've really stretch the limits in scale (ball vs paddle @ high velocity), it still needs a lot of work.
if you can stomach it,
turn your speakers off and
watch this. since it's on autopilot, the ball should never hit the bottom wall and turn red (while it obviously does, at times). and, if anyone is wondering, that
is actual (unnecessary) speed
at the upper end of "normal" gameplay speeds, it's still only 99% solid but i think i can get to 100 if i ever revisit the project (which, obviously, this has me considering).