Lloyd Chambers has tests of Pentax 645D, Leica S2 and Hasselblad H4D on his DAP site. There are comparisons with Nikon/Canon equipment. It's a pay site but worth the cost.http://www.diglloyd.com/dap/index.html
Imaging resource has samples from the Pentax645D, some available as "raw". My summary of some findings is here: http://echophoto.dnsalias.net/ekr/index.php/photoarticles/51-a-closer-look-at-pentax-645d-image-quality
Regarding the DR aspect DxO-mark has good data.
I have seen very little downloadable images for comparing DR, but Lloyd Chambers (Diglloyd) made some raw images available for download from the Leica S2, and what I could see the S2 was inferior to the Nikon D3X on shadow detail.
On the other hand Phase One has also some images available for download. They have a high contrast image with a bridge from Melburn? photographed with both the Canon 1DsIII and the Phase One P65+. Here the P65+ outperforms the Canon 1DsIII by a wide margin with regard to DR.
The above conclusions are based on my own raw-processing.
Just a couple of more points.
- A larger sensor collects more photons and will be able to hold more electron charges. Photographic noise in digital images is mostly caused by the natural variation of the number of incident photons, called shot noise. Doubling the area will reduce the shot noise by a factor of 1.4.
- A camera having good DR would also have good high ISO performance.
- On the other hand, a larger format with top class optics would increase MTF for a given feature size, giving better "microcontrast" and better resolution of texture in the shadow areas. This is also an area there better lenses may show significant advantage. The Leica S2 lenses are said to be excellent designs, but I'd guess that some of the Phase One lenses are truly excellent, too.
In short, although I share your opinion on the value of lab testing I have seen enough evidence to convince me of the benefits of modern MF equipment. If it is worth the cost is an other issue, which is pretty much person dependent.
Those people tend to forget that there are scientific methods of determining what you can see and what you can hear - not involving a single MTF measurement. The same tools are used by the food/drink industry to establish what wine people prefer, and if they can sense the difference between two wines.
If you are served two glasses of wine multiple times over a large time period and cannot sense any difference at all, it is probably not a question of "a real quality, not yet formalized", but a case of over-confidence in ones own senses coupled with (usually) knowledge of brand, feelings etc.
To venture a little less off topic:
How do you think along these lines when it comes to photography equipment? Two different cameras will typically be "different". They may have different weight, different size, different button layouts, and different color balance. This is trivial and may or may not matter for the user. Many users seems to be interested in the image quality in isolation (i.e. how good the image files out of the camera looks, or how good "optimally"post-processed files will look for certain camera settings. The fact that some cameras applies more noise reduction, while others have more sharpness seems irrelevant if one can emulate one with the other using reasonable photoshop skills and time.
I think it would be intereseting to have some tool (based on dcraw?) that could take two raw files from two different cameras of the same scene, and apply development parameters that make one look like the other. How different would cameras be then? I sometimes see two side-by-side images that are clearly somewhat different in exposure and white-balance presented as proof that one camera is inherently better than the other. Or two images with radically different tonemapping presented as proof that one camera has better dynamic range than the other.