The design intent for the AP servos is that they come on with the master and stay on throughout the flight. They are designed to survive engine cranking and other spikes at startup and shutdown. As mentioned above, if they are off when you have your EFIS on, you'll get errors, but this isn't a problem.
We do recommend that they have a pilot accessable switch or circuit breaker so that in a whole cascade of failures, you could turn them off if needed. The servos require constant communication with the EFIS to stay engaged, so they will disengage immediately if the EFIS ever fails, or if a whole host of other things occurs. We've never heard of a servo engaging itself or failing to disengage when it should, and even if it did, the design of a stepper servo means you can override it and fly the airplane.
As a company, Dynon doesn't see an advantage to a user turning them on and off independent of the master switch. Worrying about them after takeoff or before landing just seems like extra workload for no real advantage. Most standalone autopilots require you to turn them on before takeoff because they calibrate on the ground and they can't be cycled in flight, and nobody worries about the AP having power when taking off or landing. No good reason that the Dynon servos should be treated differently. But if this gives you more confidence to power cycle them manually at different phases of flight, the system supports this.