You want the longest orbit that will still hit the points you want to hit, so your final satellites don't need a lot of delta-v to circularize. Scott (in the video above) could have done a 2 hour orbit instead, but that would require more delta-V for each satellite, and the periapsis of a 2-hour orbit would probably be inside the atmosphere. The formula is like this, where P is the rotational period of the planet and n is the number of satellites: P(n-1)/n. So for Kerbin's 6 hour rotation and Scott's three satellites, you get four hours for the parking orbit. Four sats would give you 4:30, five gives 4:48, six gives five hours even and so on. For two sats, I'd say set up like you were doing four and then skip an orbit, if you try to make a three-hour orbit you'll probably crash. In fact, now I think about it, if you set up for twice as many satellites and just skip launching every other orbit, you can save even more fuel on the probes. Say you launched for a 5.5 hour orbit, which would give you 12 launch windows, but then launched a probe on every other orbit. That makes six probes, but each one has less dV requirement. It just takes more waiting. Also, if you can't get the period of the parking orbit perfect on the first try, don't launch the probes until you get it perfect, all the rest of the timing is dependent on your orbit at the time of the first launch.