Upload & Sell: Off
Rayleigh at F2.8 leaves 0.15 in MTF. A resolution just 5% higher than Rayleigh leaves 0.00, i.e 0% contrast.
The best (commercial) lens I've ever measured was the Coastal 60mm UVIR, and even that had a loss-factor of about 1.05, meaning that it could (just barely) resolve detail at Rayleigh (<0.05 contrast, barely sharp enough to sharpen into visibility). Anything above this is physically impossible unless you use phased information like in a telescope array (a combination of multiple telescopes to get one image).
So lets say the Canon 70-200IS2 is as good (aberration free) as the CO lens, then you get a resulting limiting aperture of F2.8*1.05 = F2.94.
Rayleigh for F2.94 green >>> 1.22*0.550µm*2.94 = 1.97µm
1.97µm = 253lp/mm
It is physically impossible for that lens to resolve anything above ~250lp/mm.
If you want some hard proof of this, look into high-end microscopy or astro, where they take their resolutions seriously. In those fields you cannot wave resolution figures around without verifying them.
The 63X Zeiss LD Plan Neofluar® microscopy objective combined with a really, really good tube lens BARELY resolves the 1.5µm target at an MRS-5 Geller standard at an NA of 0.19 (~F2.6). The resulting MTF is 0.07 @ 1.5µm (the Pentax Q has a pixel pitch of 1.55µm). That LD Plan lens has aberrations at least an order of magnitude smaller than the 70-200...
I don't doubt that you get very good detail out of a good lens with the Q, but the Bayer system only needs 1.4x pixel pitch to start creating fictitious detail. To get reality-related detail you have to stay above this, no matter how good your interpolation algorithm is.
1.4*1.55µm = 2.17µm = 230lp/mm = the testing limit for the Pentax Q.
I would actually think that a medium-quality F2.0 lens would resolve more than that (at very low contrasts, but still possible to sharpen). F2.8? at very low contrast. F2.8 with 1.4x TC = F4.0? Definitely not. The image might look sharp, but the real detail resolution at real pixel level isn't there. It might be close, but not there.
Determining the resolution limit with a bayer-based sensor is really hard when you approach the upper limit of the system, so the error margins for "visual inspection" gets up into the "not really reliable at any level" when you try.