On the first link, it's no contest. S5 camera is clear winner with better details, sharpness. Z2 images are often too noisy.
I agree that the Z2 has performed badly, but all are OK at "web size", and conversely none stands up to viewing full size - the noise reduction has removed a lot of detail when viewed like that.
As a little experiment, I downloaded a few of these and rescaled them to the size of the M8 image (or the long axis to the M8 size, in the case of the Z2 with it's 4x3 ratio), to try to get an impression of what the real differences in detail might be. The results were mixed. In the crossroads picture the S5 image is clearly the best - most detail and best sharpness - and the Z2 the weakest. But in the statue image the S5 is the noisiest and least detailed even after downscaling. Of course this does assume no technique failures on the part of the photographer, but the problem in that image looks like noise/noise reduction rather than blur. Maybe light was lower, or maybe a lower contrast subject was more challenging for it.
I'm actually more curious about daylight images, because to actually exploit the 16MP resolution in a sensor that size is going to put a lot of demands on the lens. In fact it's very close to being diffraction limited: for an F2.2 lens the Airey disk (diffraction spot due to the lens aperture) is going to be 2-3 microns wide (purple/blue -> red light). The pixel width is 1.12 microns. So with a
perfect lens the pixel size is smaller than the spot size due to diffraction, i.e. you are oversampling. Oversampling does bring a gain in resolution up to a point, but even with a perfect lens the optical resolution is already being limited by diffraction rather than the pixel size. I'm sure the camera designers know this (the parameters are so much on the limit that I'm sure it's not coincidence), but am curious to see whether they were able to (consistently) produce a lens that was up to this.
(The flip side, of course, is that unless the lens is
very good they could have gained on noise with
no loss of resolution by using fewer pixels.)