In the idealised case, with no loss of energy, in contradiction to Pete, 1.
The reasoning being that the speed of the topple depends on the potential energy in the erect domino, plus whatever surplus energy had been supplied at the start. So with perfect dominoes in a perfect vacuum in a perfect shape on a perfect surface, in a perfect topple (no tumbling because of that bullet shifting the domino that it hit etc) the speed of the first domino would be the speed of the last domino.
In fact, if the excess energy in each domino were conserved as well, the topple speed could even increase indefinitely with each topple, which was your item 4, I think. In fact the same would happen no matter how gently you started the topple. In real life of course, long before you approached light speed, the dominoes would begin to break, absorbing energy and raining on the parade. But if you could avaoid that, you could base a linear accelarator on toppling dominoes, or even a circular accelerator if you could replace the dominoes fast enough.
But never mind that, I am quite sure that you have had more adventurous proposals for 4...
But guess what?
In real life such perfection is not easy to arrange!
Accordingly at each impact some energy would be lost to friction, slippage, poor placement, air resistance and so on. In practice there always would be a certain loss of energy and in fact, statistically there is essentially a certainty of some deviation sooner or later interrupting the chain. But ignoring that, the process would go on indefinitely, with the speed of topple decaying rapidly (probably three or four topples) to the same constant as if you had toppled the first one by removing a support and let it topple passively.
Even that would take some impressive arranging. That sort of thing is not as easy as it looks.
In other words, what Pete said: 2!
(Note that it does not matter whether that is a factorial sign or an exclamation.)
