I don't think you're hearing me. Shooting a 50/1.4 on 135 system at max aperture is not going to give you an optimum result for the reasons I have explained. Just prior to making the switch to Oympus I did use the Sigma 35/1.4 and yes, it is a very good lens, but it is still not as sharp wide open as it is when stopped down. Especially when using unreliable phase detect auto focus on a DSLR. So in practical terms you need to stop it down to at least f/2.8 which more or less gives you the same depth of field that you would get from an Olympus 17mm f/1.8 lens. However, the Olympus lens is immediately sharp at f/1.8, so not only are you getting shallow depth of field, but you are also allowing enough light into your camera to get a shutter speed that ensures a sharp image. In practise this is much more useful than having bragging rights to a larger sensor with shallower depth of field.
I think you are neglecting to mention one very important detail. Using a smaller sensor with an f/1.8 lens is not going to give you a better signal-to-noise ratio (SNR) than a 36x24 mm sensor at f/2.8.
With modern sensors (say, comparing an OLYMPUS OM-D E-M1 to a D810) with low read-out noise, your SNR is mostly limited by photon shot noise. Photon shot noise scales directly proportional to sensor size, since we can assume that for a constant amount of incoming light, the larger sensor will capture more photons per unit time. Comparing apples to apples requires that we compare the resulting SNR at the same display size for both sensor sizes, i.e., we rescale both images to, say, 8 MP. The net effect is that each pixel in the 8 MP image coming from the 36x24 mm sensor is the average of incoming photons over an area that is 4x greater than the corresponding sensor area coming from the m4/3 sensor (18x13.5 mm assumed). This averaging improves the SNR.
In terms of exposure, this means that given identical sensor technology, you can decrease your exposure by two stops on the D810 and still obtain the same SNR as the E-M1. You can check the DxOMark SNR values yourself. For example, E-M1 at ISO 200 yields SNR of 37.4 dB, and D810 at ISO 800 yields an SNR of 36.8 (yes, I know 0.6 dB is actually a fairly large difference, but it is impressively close given all the variables in the comparison).
So if your Sigma 35 mm f/1.4 is stopped down to f/2.8 on a D810, you will still have less DOF and you will be able to use double the shutter speed while maintaining the same SNR. (This example is just to tie in with your settings quoted above)
Using VWDOF we can calculate that the roughly equivalent DOF will be attained when we compare the m4/3 at f/1.8 to the 36x24 mm camera at f/4 at a focusing distance of 1.8 m. Since f/4 is roughly 2 stops above f/1.8, we can use ISO 800 on the D810 to keep the shutter speed the same as the E-M1 @ f/1.8.
To summarise: you cannot work around the physical limitations imposed by photon shot noise. The upshot is that given equivalent sensor tech, the SNR of m4/3 at f/1.8 ISO 200 will be virtually identical to that of a D810 system at f/4 ISO 800, with both at the same shutter speed,
and the same DOF.
There are many ways in which to skew this comparison, e.g., comparing images from both cameras at 100% size on a monitor, but the reality is that the same result can be obtained with either system.
What the m4/3 system cannot do is obtain a shallower DOF; whether that is desirable is a matter of taste. Yes, it is harder to obtain accurate focus at such shallow DOFs, but with careful AF adjustment you can obtain accurate results. Also, we have not really specified what type of photography is involved --- if the subject is static, then DSLRs offer a passable contrast detection AF option (but yes, it is not as responsive as the AF systems in mirrorless systems).
But please do not take my contribution to this thread as an attack on mirrorless systems in general. Given enough time, the mirrorless systems will match DSLRs in the remaining categories (dynamic focus performance), low EVF lag, etc. The reality is that very few people actively use their DSLRs in settings that require the sophisticated tracking AF offered by top-end DSLRs like the 1Dx or D4s.
I just wanted to point out that in order for the discussion to progress, the comparisons between systems must be made in a fair way.
