"what he does in the "repstrap" model is somewhat irrelevant, no?"
Yes, its very relevant. Please understand this is a very complex project. The vacuum system along with the electron gun systems are not at all simlpe. This isn't a "hot glue gun" moved around by stepper motors salvaged from old printers and mounted on wood. The "repstrap" model will have three main systems: Vacuum, Electron Gun and Motion control. The motion control system is the easiest part as it can use standard reprap designs. The vacuum system and electron gun system are both complex and require quite a bit of research, effort, time, money and trial-and-error. Even after the electron gun is built, you need to focus and deflect the beam which again will require a lot of research and development. Electron beam welding machines cost half a million USD for a small machine and millions for larger machines. There is a reason they cost so much. Also many companies have spent a lot of time and money over the decades researching better gun and column designs.
I do not want to sound negative or discourage anyone from attempting to build a metal printing reprap. However, I do want to be honest and ensure no one has false ideas of how to get one working. Its naive to assume vacuum is easy or that electron gun design is easy without ever working with either of those systems. Don't underestimate the complexity of this project, its the most complex of any reprap variant.
"I mean, he needs to get to the vacuum he needs, but once he has a running EBM repstrap, 3D printing a roots blower or a turbomolecular pump should be pretty easy, I would think, and then all future EBM printers would be able to use a *cheap* roots blower."
I am going to say no, it is not as easy as you think, unless the printer has a tolerance down to the thousandth of an inch or better (~0.01mm). Realistically, vacuum is a fickle beast. The slightest scratch or imperfection will render a pump or sealing surface useless. Everything in a vacuum pump or chamber needs to be as smooth as possible otherwise you are trapping dirt and moisture in the cracks and crevices which slow or even prevent the chamber from reaching a decent vacuum level. Any blind screw holes in a vacuum chamber will also trap air and moisture. Screws in blind holes are either drilled through with a small diameter bit (works for larger diameter screws) or have a groove or two cut parallel to the length to allow trapped air to escape. Those little pockets of trapped air which leak out slowly cause what are known as virtual leaks; leaks that cannot be traced to any physical part of the vacuum chamber but are instead within the chamber itself. Another thing to remember about vacuum is there are two modes of flow, viscous flow and molecular flow. Viscous flow occurs at higher pressures from atmosphere to the mid -2's. Once you reach the mid-low -2's Torr, you enter molecular flow. Molecular flow is why you want smooth surfaces as the air molecules are no longer under any pressure. They freely bounce around the chamber under their own energy (heat). You want them to bounce around as little as possible and bounce into the pumps.
Refrence these two pages:
http://www.chem.hope.edu/~polik/Chem345-2005/vacuumtechniques.htm
http://www.vacuumlab.com/
Building your own turbo pump is another can or worms, these are high precision machines. The blades are thin and need to be well balanced as they will need to spin at 50-100k RPM. The blades are better off being laser cut and formed or stamped. Plus you need a brushless motor with special bearings so it can spin very fast. Custom built electronics would be needed for an inverter as 50/60Hz mains frequency is not even close to the speed you need to run a brushless at those speeds. You need 1600 - 2400Hz or more depending on the number of poles in the motor. Refrence:
http://en.wikipedia.org/wiki/Turbomolecular_pump#Practical_considerations
I will say that a diffusion pump may be pretty easy to build depending on your available tools. Its simply a metal cylinder wrapped with a cooling tube and a heater plate on the bottom. Inside are some sheet metal parts for the oil vapor. You could build one in a machine shop equipped with a lathe and a TIG welder and some stainless steel tubing. The "christmas tree" inside the pump can be made from aluminum tube and sheet metal, most christmas trees are.
Bottom line: The vacuum system is the biggest and most expensive hurdle to overcome. You can save money here and there but in the end the pumps are your biggest expense.
Its a long road to getting a metal Reprap to self replicate. I would not even hesitate to say it might not be possible without machining and/or grinding the printed parts further to obtain the necessary tolerances. Think of it like casting parts, you can get close to the desired shape but there is always some machining to do afterwards to gets thing smooth, square and balanced. But lets not even think about that until we get a prototype shooting a beam into some metal powder.
A long winded post but I assure you this is a very involved project and I commend everyone involved for putting in the time, effort and research. Hopefully one day we can have a self replicating EB fabrication machine that can also drill and EB weld as well.