You'd think so. I had problems with it updating while the ball was technically in the floor. I had it detect when it hit the floor by when it passed the frame (obviously) but then it would switch the vector and update the position but the resulting position was still under the floor and it would do it again. So the ball would bounce like 4 times then get stuck in the floor.
<script type="text/javascript">alert("remember when scripting attacks worked?");</script>
There are ways of dealing with that. One is to have a finer resolution of time for the physics simulation than the framerate. If you only have one moving object and static obstacles, though, you can set it up to simulate either to the time limit, or until the next transition (hit an object, rolling up instead of down, etc.), then keep doing it untill the required time (the frame interval) is elapsed; this way you will have no strange results, but it doesn't scale well to complicated simulations.
What I ended up doing was adding a timer to the simulation so that when the ball hit the "ground" at least 2 frames went by before it would preform another collision check. This actually proved to work very well and even through extensive testing I never found a problem with it.
<script type="text/javascript">alert("remember when scripting attacks worked?");</script>