Hey, I've encountered a rather awkward tunneling issue while implementing this.
Let's say the situation is as follows:
(apologies for terrible scaling : P)
Red is the dynamic object's current position. It is moving 4 pixels down and 10 to the right next frame.
Black are static objects.
Purple is the impact position, Blue is the position by the end of the frame, according to swept.
As you can see, since swept does not check the rebound, there is a chance for objects to tunnel on such impacts. My first thought was to repeat collision detection on every rebound, however, this results in nearly infinite loops at low velocities. So, next I tried using basic AABB at lower velocities, but this too failed when sliding at >swept threshold velocities. I've been stuck on this for almost two weeks now - does anyone else have any possible solutions to this issue?