When you say "powerful enough", consider that any roll torque would be accumulating over time -- even a penny will eventually turn into a huge fortune with compound interest, given enough time.
Yes, I didn't think about it.
Presumably, highly-optimized EMdrives would have frustrums made of materials offering highest Q - which would normally be superconductors. Hopefully, some combination of thermoelectric and radiative cooling would guarantee the superconductors can be reliably kept cold for a decades-long voyage, to ensure continuous EMdrive operation for the duration of the trip.
If your EMdrive broke down part-way during the trip, then you could be headed for your destination without any means to slow down. You'd either hit it, or overshoot it.
If EMdrive is to be reliable for a long-distance/interstellar journey, then all of its components have to be reliable, and all of its supporting systems (power supply) have to be reliable.
What kind of reactor design would be most reliable for long duration voyages?
I agree with you, reliability of EmDrive components is paramount. In an early phase, with first generation e.m engines, it would be a wise choice to add a fair amount of spare parts to the payload.
I think, as rfmguy already said, that RTG is a good choice but, I add, for slow but steady thrust values comparable with actual EmDrive. If we talk about future manned interplanetary/interstellar starships instead there'll be the need of something capable of generating hundreds of thousands (millions?) of amperes to power the main thruster (I'm thinking to a steady 1g acceleration). This amount could be probably reduced, I suppose, with technological advancements of the engine: better materials, better configurations, a deeper understanding of the principles etc..
Regarding forces arising from course correction - just as you said with the fliparound, if the magnitude of that rotational torque/force is not too large, and if the main thruster could be shut off in the meantime (or at least throttled back), it could help in keeping loads tolerable. In space, there's lots of room to maneuver.
The idea of keeping the main thruster always active is to have a constant illusion of gravity due to the 1g acceleration. However for probes and less sci-fi future spaceships with far less power, that is astronauts still experiencing microgravity under constant acceleration, your solution is effective and easily applicable.
.. Solar panels to jettison near Jupiter, RTG takes over from there at reduced power level.
Why to jettison the solar panels? The e.m drive (provided it's powerful enough) wouldn't have problems to carry extra weight, as it has no fuel constraints.
F=MA
A=F/M
Lower mass = faster acceleration
Does EMdrive have to be placed on the back end of a spacecraft? Can it be placed on the front end of a spacecraft, like "front-wheel drive"?
Whether or not it's a "Q-thruster", does EMdrive produce any appreciable "exhaust" that people in the vicinity of its resonant cavity would have to worry about?
What are the main safety precautions/considerations required for people in relation to EMdrive?
Does higher Q-value (eg. superconducting cavity/frustrum) minimize any radiation leakage hazard?
Does EMdrive have to be placed on the back end of a spacecraft? Can it be placed on the front end of a spacecraft, like "front-wheel drive"?
Whether or not it's a "Q-thruster", does EMdrive produce any appreciable "exhaust" that people in the vicinity of its resonant cavity would have to worry about?
What are the main safety precautions/considerations required for people in relation to EMdrive?
Does higher Q-value (eg. superconducting cavity/frustrum) minimize any radiation leakage hazard?
So a Valkyrie style tractor spacecraft, with a tensegrity spine? (a recent example of a tractor configuration being the ship in the movie Avatar).
If there is an "exhaust" issue, using a slight thruster cant from the primary axis shouldn't hurt too much if the spine is long enough.
But if there is no exhaust issue, then there would be the tradeoff between larger individual thrusters, and distributing the engines along a tractor configuration spine to cut spine weight (if the cargo can be modularized to be spread out on the spine).
So a stack of T's versus a tall T shape then, assuming you didn't centralized the thruster on the spine itself (differential thrust from thruster pairs on arms at the top of the T seems attractive).
Extreme T example is a thruster bar with a very long rollout solar array trailing (perhaps with inflatable/rigidizable tubes to stiffen the solar panel after deployment). Make the thruster bar cylindrical and you could ostensibly wrap the solar array around it like a roll tube, then inflate deploy out like a party favor.
Space drive has actually 2 models
1-EM drive is nonrelativist and beam "photons" are quasi particles with accelaration of order a~10ēģ m/sē
It uses Newton's third law for its thrust !!
2-Alcubierre warp drive using Emergent gravity techology not the Quantum vacuum plasma technology of Harold "Sonny" White
From Mike McCulloch's blog, where I made this point...
One thing we should discuss is the power supply needs for the 1 gee space-craft and the EM-Drive parameters needed for that performance. Say we have a 1,000 ton spaceship and we accelerate at 1 gee. Thus 10 MN thrust is required. At Eagleworks' observed performance of 1.2 millinewtons per kilowatt, the power to get 10 MN thrust is 8.3 TW, which is probably unrealistic for a 1,000 ton ship to supply. We might be able to do it with a laser bank beaming at a high-efficiency collector, but it's still herculean by modern standards (we're just starting to make sustained 100 kW lasers.) It's still much better than the laser power needed for a purely photonic drive, some 3 petawatts, but it's a long way from our state of the art.
Of course the performance increases with the Q-factor. Tune the cavity and make it superconducting. If we take the NASA EM-Drives and pump the Q factor to ~30 million, then about 2 GW power is needed for the sustained 1 gee thrust. A fast-spectrum reactor with a thermal output of ~ 6 GW and ~35 % thermal conversion efficiency would be a first pass design to supply the power. Assuming ~100% burn-up the fuel used over 20 years masses 4.2 tons. If the reactor mass was limited to ~200 tons, then it'd need to supply power at 10 kWe/kg of reactor mass, which is very high performance. A gas-core or magnetic collimator fission-fragment reactor might be up to the job, but both are somewhat futuristic.
Fusion reactors are presently *hoped* by propulsion engineers to be able to supply power in the 100 kWe to 1 MWe range, though no working fusion power-reactor has yet been demonstrated. As the joke goes, fusion has been "just 20 years away" for the past 60 years and always will be...
I wonder just how high a superconducting EM-Drive's Q-value can go?
Addendum: I've read of Q factors in the billions. Thus 20 MW would then be sufficient for a 1,000 ton 1-gee spacecraft. Much easier to supply with a mobile fission reactor.
I believe we should re-imagine our concept of spaceship (and probe)
First of all, a spacecraft equipped with an EM engine can thrust as along as the power source is active. This means it can not only visit various celestial bodies in a single mission but also that it can reach speeds that are unattainable with ion/chemical propulsion. It can basically accelerate up to half way from destination, then turn 180° and thrust in opposite direction to slow down.
The best solution to power the engine is a nuclear reactor (and it's even fundamental for deep space missions): I'm talking with PNN in mind, which requires a great amount of current. I think solar panels would be an excellent solution for early EM probes with low thrust values and satellites.
Until we won't have a spacecraft specifically designed around EmDrive (or other EM thrusters like Cannae or PNN) I think our current models will be retrofitted to mount the frustum(s) or tiles (PNN).
I can't help but it comes to my mind an imagine of the Space Shuttle equipped with an array of frustums 
I depict this near-future like the hybrid old-new vehicles I've seen in Independence Day Resurgence (but that's daydreaming).
E.M technology would make excellent position thrusters to rotate a spacecraft around all its axes.
If we imagine that tomorrow morning EM propulsion will be used for space application then we can speculate that in 10/20 years the technology will be mature enough to generate massive thrusts, even capable of generating more than 1 g of acceleration. Then a spacecraft will probably levitate and gain orbit in a little time.
We'll also probably laugh at how wrong all sci-fi spaceships have been depicted so far: basically sea ships design transposed in space. With such design if the ships accelerates the crew (and everything is not bolted on the hull) is pushed toward the stern. With 1 g acceleration the best design is.. something similar to a skyscraper 
Personally I think that flying saucer design is good too (except for the fact that during space flight it offers a big impact surface to debris like rocks).
Also interstellar travel become instantly far closer than with chemical/ion/antimatter propulsion (basically everything Newtonian).
I invite you to read these two articles from my blog (I link them because as far as I know nothing similar exists on the web)
Sailing across the solar system and beyond/
As if one wasn't enough PNN breaks another law/
Those are written for PNN but I believe they're also valid for EmDrive
The PNN stuff seems like utter nonsense. Their video showed absolutely nothing.
