Will need Dan to check if the "snap to value" continuing to execute in design mode is a bug or not. Just sort of is.
I understand the Start and Stop button behavior, but the Halt button looks like it might rely on a bug I thought was stamped out previously. When a subcontrol has more than one super, it is not supposed to feedback. But in this case, the Halt button appears to be getting feedback from the Driver, and when pressed, sending that same value back to the Driver. So if there is a bug, seems that it is useful in this case.
And that same feedback seems to be at play with the start button. At the end of the first cycle, the start button goes off, then stays off for remaining cycles.
As far as freezing MDP, we are sort of outside design conditions here. MDP2 was designed to prevent feedback, but you have found a way around that, in a useful way. But you can force crashes - try setting the start cycle time below 500 msec, and it will lock up. But I think this LFO behavior is useful, so we want to keep it, just gotta use it carefully.