If you had read their testing methodology, you would see that they vary focus to account for field curvature.
They don't. Some of their tests have separate MTF charts for 'best at any given point' sharpness, but most don't. Read these two articles for further information:
https://www.lensrentals.com/blog/2017/11/testing-lenses-finding-the-best-average-focus-point/https://www.lensrentals.com/blog/2017/11/testing-lenses-best-individual-focus-mtf-curves/In any case, high-quality lenses of this type (medium telephoto) tend to have very low field curvature, and I would expect that to apply to these lenses.
When you're measuring MTFs at 50lp/mm, every small misfocus counts. Small focus errors which would probably not be noticeable in a typical photo (with certain exceptions) show up as big differences in measured numbers.
Case in point - try taking a photo of a night sky, with discrete, individual stars. Your focus needs to be perfect - even the slightest misfocus turns the pinpoint stars into mushy blobs. Use the same degree of misfocus in typical landscape photo and you'd never notice.
The optical bench measures pinpoint dots of light - just like stars. Tiny focus errors turn into big differences in measured MTF. Look at the 'average MTF' (focused on centre) vs 'best focus MTF' (focused on point being tested) for the Sigma Cine 85/1.5
here. This is a lens with a very flat field of focus, yet there is a dramatic difference between the average and best MTFs measured.
As to focus at infinity for the tests, I don't see where you get that from. It is impossible to shoot a test target at literal infinity.
They use an optical bench to generate an infinity target. Incoming light is parallel to the axis of the lens, equating to an infinite target distance.
We should all note that this test is a very small part of the overall story, since they are only testing at the widest aperture, and it is often the case that lens performance will improve as the aperture becomes smaller (until the point that diffraction starts to progressively degrade quality). However, if a lens performs really well at its widest aperture, it is clearly a superior lens.
Not necessarily. Unless there is a clear-cut difference where one lens is better than another in almost every measurement at every point, use case matters a great deal.
Consider two lenses - Lens A, a 500mm f/4 lens which is extremely sharp in the centre, but drops off significantly towards the edges, and Lens B, a 500mm f/4 lens which isn't quite as sharp in the centre, but which maintains its sharpness right out to the corners. Lens A will give you an extremely sharp centre, with softer corners, while Lens B will be moderately sharp throughout. Not an uncommon situation - but which is the better lens?
Remember, these are large supertele lenses, likely being used for sports and wildlife, often with teleconverters (which essentially take the middle portion of the image circle and magnify it). It will be rare for the entire field of view to be in focus, and rare for the point of the focus to be in the outer parts of the image circle. When using a teleconverter, the outer parts of the image circle won't even be in the frame at all. In this case, Lens A, with the sharp centre and average corners, is probably a better lens - in most cases, the edges will be out of focus anyway, but you want extreme sharpness in the central portion for fur/feather/scale detail and for teleconverters. Lens B will just give you some very sharp out-of-focus details in corners.
Now take the same two lenses, with exactly the same MTF charts, but make them 24-105mm f/4 lenses instead. Which one is better now?
These are now all-purpose, relatively slow wide-to-medium-telephoto lenses, being used to shoot everything from portraits to landscapes (possibly even some pseudo-macros, depending on maximum magnification), likely with more of a focus on travel and landscape photography than being dedicated portrait lenses. Lens A still gives you a great centre and so-so corners, while Lens B gives you good, consistent sharpness corner-to-corner. Out of these two lenses, Lens B is probably the better one. Same MTF charts as the 500mm f/4 example, but the different purpose of the lens makes the priorities and desired characteristics different.
Now consider a pair of 70-200 f/2.8 lenses. Lens C is super-sharp in the centre at f/2.8, with so-so corners, which improve to being sharp corner-to-corner at f/8, while Lens D is sharp across the frame at f/2.8 (without being as sharp as lens C in the centre), with slight-to-moderate general improvement across the frame at f/8. Which lens is better? Probably Lens C. At f/2.8, it's probably being used for portraits or similar shots, where focus is on a specific person or feature, rather than the scene as a whole. The corners are likely going to be out of focus anyway, so Lens D's across-the-frame sharpness adds nothing there. When shooting a landscape or other scene where everything is in focus, the lenses will probably be stopped down to f/8 or narrower. Both lenses likely perform similarly here, but Lens C's sharper centre when wide open gives it the edge due to versatility (being better for portraits while being just as good for landscapes).
Now make them 14-24mm f/2.8 lenses instead, with the same MTFs. Now Lens D is probably better, because every shot is likely to be in focus corner-to-corner anyway, with wider f-stops being used more for shutter speed and low-light situations than for DOF control. Lens C will give you mushy corners and be all but unusable at wide aperture, while Lens D will be much more usable without having to stop down. Compare and contrast Canon's 16-35 f/2.8L II (not the current Mk III version) with Nikon's contemporaneous 14-24 f/2.8 - Canon's was sharper in the centre but had mushy corners, while Nikon's was better across the frame. Nikon's was widely regarded as the far better lens.
MTFs are meaningless in isolation, without considering the overall purpose of the lens - including lenses with different purposes at each end of the zoom range or at different f-stops. What constitutes a good MTF curve varies greatly depending on intended use - do you need corner-to-corner sharpness, or a super-sharp centre? A lens with a 'worse'-looking ETF than another may actually turn out to be a better lens once you consider what it's actually going to be used for, even without considering non-optical characteristics such as AF or weather sealing.