Again, the speculation that it is IoT devices is unfortunately just that. However massive compromise of internet connected embedded device is not new: http://internetcensus2012.bitbucket.org/paper.html
TLDR: Now deleted forum comment by a scientist says that the paper "Measurement of Impulsive Thrust from a Closed Radio Frequency Cavity in Vacuum" has passed peer review and will be published in AIAA Journal of Propulsion and Power. A line of text was included that may be from the paper itself: "Thrust data in mode shape TM212 at less than 8106 Torr environment, from forward, reverse and null tests suggests that the system is consistently performing with a thrust to power ratio of 1.2 +/- 0.1 mN/Kw ()"
Depends on how well does it scale; here is some back of the envelope math:
If we take a Topaz reactor design and build a spacecraft say 10 times bigger than new horizons this gives us a mass of 5000 Kg.
We'll dedicated 2000 of those to the reactor and fuel, and 2000 of those to the instrumentation (this alone is 5 times the dry mass of New Horizons) and 1000 KG for the platform/chassis.
A Topaz reactor weighs around 300KG, let's say ours weights 500 because we want to be able to swap fuel in and out and we want to make sure it generates 5KW of power.
It can operate for about 5 years on 15 KG (12 in the original design) of fuel which in this case is UO2 gas (spent fuel might be used as propellant so we might be able to have a secondary Ion drive cycle to maximize the thrust we get but won't use that in our calculations) this means we have fuel for 100 years.
4 of our 5KW would be dedicated to our propulsion system which would give us a ~4.8N of force which at a mass of 5000KG means we get 0.00000096 m/s2 of acceleration.
Given V=V0 + at after one year or 31622400 seconds accelerating at 0.00000096 m/s2 we'll have a velocity of 30.357504 m/s after 50 years our velocity would be 1517 m/s or about 10% of Voyager 1 and 2.
So yeah not so great unless we can scale it much much better, that said if we do figure out the physics behind it we might be able to design more efficient emitters with considerable better thrust to power ratio.
