_______
PS: concerning the axionic dark matter explanation we still have to hear from @frobnicat on his warning that it maybe several orders of magnitude off the measurements. I also have a concern due to some of the estimates made for the density of dark matter expected around our planet (being too low to explain the measurements), but that is an unsettled area of research.
yes yes, me and my big mouth
Let's be clear that it's about the use of naturally occurring dark matter (be it light Axions rather than Wimps if the involved frequencies are indicative) and not about the production or use of dark matter considered in an "empty dark field" (gosh) that is at its rest state, lower point energy. The question were : is it worth to use this naturally occurring flow for momentum or energy? (which is the same as momentum = energy / c if we take the photon rocket as a proven baseline of "best spent mass_energy")
In fact I then tried a back of the envelope arithmetic like that :
mass flow = density * average speed * device cross section
kg/s = kg/m^3 * m/s * m^2
a few assumptions there : the dark flow is homogeneous in density and speed (so I guess this discards hot dark matter). I'm also assuming 100% of dark mass going through the device is "used" somehow (which for a weakly interacting particle would be a rare manifestation of goodwill).
Then I went other equations, like F=mflow*Vej and Pow=0.5mflow*Vej^2, going through the energy it takes to push on such a flow to get the reported thrusts (powered "propeller" device, not just passive sailing, which I believe would have seen quite a strong dependence on sidereal time, no thrust reversal with reversal of device...) and with the idea of seeing if the necessary velocity of ejection is indeed higher than the average velocity of flow that brings mass (therefore validating some of the "propeller&slow wind" hypothesis). Sorry for the convoluted line of reasoning and slow brain tonight.
At about 1Gev/cm^3 (as seen on this
seemingly optimistic paper, maybe more optimistic is possible) and .01m² csection and 250km/s dark flow velocity that is mflow=1e9*
1.8e-36/1e-6 * 2.5e5 * 1e-2 = 4.5e-18 kg/s.
With about 45*µN thrusters that yields Vej = F / mflow = 4.5e-5 / 4.5e-18 = 1e12m/s

Pow = .5 mflow Vej² =

= .5 * 4.5e-18 * 1e24 = 2.25 e6 = 2.25MW hence the 6 orders of magnitude boast (more like 5 actually with those numbers) when comparing to 20W power.
Since thing is relativistic I were humbled, and wrong to be starting with Newton. Seems you don't have the patience to wait a month before I gather my spirits around SR so...
The best use that could be made of recovering this mass flow : converting it to photons, sent collimated and push on expelled momentum. This is, standard, please feel free to disagree. Then just see the power equivalent of the dark mass flow and compare it with what we would get from that as a photon rocket :
darkPow = mflow c² = 4.5e-18 9e16 = .4 W and that would push at F=darkPow/c = .4/3e8 = 1.3 nN (nano Newton). Dark flow average velocity times mass flow also would impart a recoil of 2.5e5*4.5e-18 about 1e-12, a pN. This would be less efficient use of this incoming mass_energy flow (and incompatible with thrust reversion).
How does it (1.3nN) scales compared to the 45µN/20W numbers above? Well I don't know what the 20W have to do in it but to scale nN to 10s of µN is 4 order of magnitude higher, that is needs 100m² of "perfect dark harvestor" cross section, given 1GeV/cm^3 and 250km/s average dark flow. Maybe a strong dispersion in dark matter particles velocities around average velocity could make up for that, how hot is hot dark matter, could it be relativistic ?
Looks like that there is not enough dark matter though to be of practical value (in explaining the reported results in classical terms) since momentum benefit of harvesting dark mass and pushing on it can't beat the momentum you'd get from harvesting the energy and sending photons, and that energy is too low. Value of the "dark field" in BSM theories in another matter entirely.