I'm trying to make up my mind on this. In a perfect plastic collision, the combined masses' speed is less than the initial mass', and that happens instantaneously, and therefore acceleration (deceleration in this case) is infinite. Of course no real world collision is a perfect plastic collision, but assuming it's overwhelmingly plastic, I can't help wondering if the velocity does not decrease on each impact just by conservation of momentum.
A perfectly inelastic collision (or, indeed, a perfectly elastic one) doesn't have involve instantaneous changes in speed; all that it specifies is the initial and final speeds, before the two objects touch and after all the processes of the collision have come to an end. It's easier to understand that if you think about what actually happens in an inelastic collision.
When two objects collide inelastically, they lose energy by damaging each other, typically by crushing or permanently deforming each other. That process doesn't have to be instantaneous. A classic example would be a collision between two cars with crumple zones; after the initial contact, the two cars keep moving as the crumple zones are crushed, but slow down as the energy of the collision is absorbed in deformation energy. Each car experiences a deceleration due to an applied force, but that deceleration doesn't have to be infinite, even in a perfectly inelastic collision. If two cars of the same weight collide at equal and opposite velocities, and both eventually come to a dead stop, then all the conditions for a perfect inelastic collision have been met.
And similarly with an elastic collision; both objects experience forces as they deform elastically, and then again as they spring back. If you work out the forces and accelerations in either case, you'll end up with the same initial and final velocities of both objects as you get if you just assume perfectly elastic, or perfectly inelastic, collisions.
So when two objects collide inelastically, they experience forces, and the details of those forces depend on the strength, shape and elasticity of the objects colliding. There's no law that specifies how quickly the collision process takes place, and clearly a longer-lasting collision, with a greater amount of damage needed to absorb the energy lost, will exert a smaller force. So, in the case of a falling object, if the force from the collision is less than the force due to gravity, overall the falling object will not actually decelerate, it will just see a reduction in acceleration.
Add to that one other point: that the 90+ floor collisions JSanderO mentions, and whose existence is almost an article of faith to some truthers, didn't actually exist as separate events. Bits of the falling mass collided with bits of the stationary mass more or less continuously, with each collision being relatively small and absorbing relatively little energy, but all of them adding to a continuous, though not particularly constant, force on whatever was falling at any given time. The result was that only the largest jolts, where a large number of these small impacts happened to combine, would have actually reduced the acceleration below zero; in fact, we only see one in the collapse of WTC1, though there are other smaller jolts where the acceleration reduces significantly.
If the towers had collapsed like a stack of planks falling on each other, then the whole missing jolt argument would have some merit. And in fact, in verinage demolitions, where for the first stage at least the collapse is that orderly and controlled, the jolt is in fact seen. But it's because the collapse was complex and uncontrolled that there was never a clean enough impact to see a jolt. And that's one of the great ironies of the missing jolt hypothesis. The absence of the jolt is actually evidence against, not for, a controlled demolition, because only a controlled process is orderly enough to produce a jolt.
Dave