What I think people aren't getting here is that a "plastic digital M" just would
not work. You can't slop together an interchangeable-lens rangefinder camera the way you can an SLR.
An SLR has to be manufactured precisely, but the precision needed is all concentrated in one simple assembly: the mirror box. You've got to locate one plane, the lens flange, relative to two other planes, the film plane and the focusing screen plane. These are all static parts; once you get them in the right relationship, they're going to stay there. You've got one moving part, the mirror, that has to be located accurately at a 45-degree angle, but you can do that with a simple mechanical stop. You need precise tooling to create the right relationships among these parts, but once the tooling is set up you don't need to do a lot of adjustments. That's why SLRs lend themselves to high-volume manufacture: you have a big investment in setting up your casting and milling operations, but once that's done you can bang out hundreds of thousands of units.
An RF camera is fundamentally different because you've got two separate, very precise subsystems that have to be synchronized to each other. The lens-to-film-plane alignment is a bit simpler because you've only got two surfaces instead of three. But then you've got a very precise, complex and completely separate mechanism, the RF/VF assembly, that has to be located accurately within itself
and with respect to the lens assembly, so that the coupling arm that connects the lens to the RF/VF lines up in exactly the right place every time. You can't do that with a molded plastic chassis and a fully-automated assembly process. You need careful assembly and expert hand calibration to mesh that complicated, precise RF/VF to the lens focusing system, and you need to mount everything on a chassis that's rigid enough that the critical relationships stay put once you've set them.
Besides, an RF requires more inherent precision than an SLR because the exquisite accuracy of its focusing system makes any slop more painfully visible. A lot of us have tried this little experiment, but if you haven't, try it now: Set up your RF camera on a tripod, point it at a nearby wall, and focus very carefully on some well-defined detail so the rangefinder is lined up exactly. Now remove the lens and stick a little piece of transparent tape on the RF cam. Put the lens back on and look through the viewfinder; if your camera is in decent adjustment at all, you'll now notice that the RF is definitely "off" slightly. That piece of tape is ludicrously thin -- I just measured one and got 0.05mm -- but it makes a noticeable difference to your RF calibration.
Now try the same thing with an SLR, sticking your piece of tape on the back of the lens flange to displace the lens by the same amount. Unless you have the eyes of an eagle, you won't notice one doggone bit of difference in how the viewfinder image looks. It's not that one type of camera is more tolerant than the other -- it's just that the RF camera makes even small differences agonizingly visible.
At this point you might be saying, "Wait, there HAVE been cheap interchangeable-lens RFs -- what about the FSU ones?" Well, part of that cheapness was in metal finish and whatnot, and part of it was an illusion created by Soviet economic policies that subsidized the real cost of goods for the sake of getting foreign currency into the country. Some of the cheaper FSU cameras may have used crude techniques to get the necessary precision -- such as those paper shims they used under the lens flanges -- but they still had to be adjusted precisely.
The other approach you can take to get some cost out of the design is to do what Contax did with the G1 and G2 -- eliminate the super-precise optical/mechanical rangefinder assembly and substitute an electronic autofocus module that could be assembled and calibrated by automatic equipment, and ditch the exactingly-calibrated coupling cam on each lens in favor of a ROM chip to store the calibration data and an electronic "chopper wheel" mechanism to handle the coupling. This approach worked quite well -- but was more limited in what lenses it could accept. And without visual confirmation of focus, a lot of people who tried it never got fully comfortable with it.
What I'm saying is that the dominant way of making cameras less expensive today is to harness automation and electronics -- but the nature of a rangefinder camera limits the degree to which automation and electronics can help.
I've said this before in another thread: If you tried to take a Leica M8 and "decontent" it to the minimum acceptable level -- less-precise rangefinder, fewer viewfinder frames, smaller sensor, more utilitarian finish, etc. -- you'd wind up with an Epson R-D 1, which is still a pretty expensive camera (and a pretty darn nice one.) I just don't think you could make a modern rangefinder camera much less expensive than that and still wind up with an acceptable product.
So, the high cost of a modern interchangeable-lens RF camera can't be helped. Most photographers are happy to take a pass and go with the more cost-effective choice of a high-volume SLR or an automatically-assembled "prosumer" camera.
The remaining tiny minority of us just have to bite the bullet and pay the higher cost for what we feel is a better way of seeing and photographing.
