Here's a Sigma SD9 curve from a QuickMTF slant-edge test, lens unknown:

Although it can measure detail beyond Nyquist up to 0.572 cy/px, that detail is false. So back to MTF, we see that MTF50 occurs at 0.282 cy/px which is OK but only on the per pixel basis. At the sensor, the resolution is 31 cy/mm, as in 31 lp/mm, not particular good compared to other cameras,
Now let's look at my Lumix DC-G9:

On a pixel basis, the 21 MP G9 compares miserably to the 3.4MP SD9: 0.157 cy/px versus 0.282 cy/px ...
... but, on a spatial basis (cy/mm), the G9 is the clear winner 47 lp/mm vs 31 cy/mm and additionally won't show as many artifacts as the SD9, if any.
From which I conclude that the better measure of resolution in this case is cy/mm aka lp/mm
Anecdote: I once sold a Panasonic Lumix m4/3 camera versus the SD9 based on cy/px, duh ...