As mentioned above, I did manage to get the rail that had the worst baked-on "gunk" soaked in citric acid for a while. However, the garbage can I used had a hole in it, and the rail pierced the thin garbage bag while I was scrubbing it under water. So, the rail isn't as clean as it probably should be, but it's a lot better. If it proves to be a problem. I'll detach the rails and re-soak them later with a dedicated apparatus (probably a sonotube concrete form with some end caps fabricated). Let's get to the pics, though.
First up, attaching the rail support beams (pictures 1 and 2). In true belt-and-suspenders fashion (redundancy for strength), I am using both a gusset and a mounting plate for this critical join. I finger-tightened the nuts, but in trying to wrestle the 7' long beams, I managed to knock them out of alignment. Fixable, but an irritant nonetheless. You can see the gap between the gusset and the bottom of the beam on the left side. I just didn't have the energy to fix that until I get an alignment tool.
Note: these are GFS8-4590 beams with the 45mm sides milled "flat" (+/- 0.1mm max deviation). I'm sure one can get better tolerances, but not for anywhere close to the same price in my experience -- if you can find someone even willing to do the job. The bigger "cost" of using the milled surface, as mentioned previously, is that Misumi won't do certain types of machining on the milled surfaces. It's a trade-off, but I'm not regretting getting the milled surfaces, merely that I didn't get the ends tapped and the corner posts counterbored.
The third picture shows the center legs installed, plus some extra structure I added to stiffen up the middle of the rails and ensure the width didn't vary too much over the length of the machine. These legs are pretty critical since they are needed to help control deflection. If I spanned the entire 2130mm, my maximum deflection with a 75kg load would be over 0.3mm. When I cut that span in half, the deflection drops to 0.04mm. I looked at several of the "commercial" (DIY+) offerings in this size/price class, and found instances where there were long spans with single extrusions, and I worried -- a lot -- about whether the deflection was going to eat away any of the supposed accuracy in the machines. Once you play around with the deflection calculators a bit, having long, unsupported spans will not look like a good idea. As it is now, I will watch what my results are to see whether I need to add more legs/structure in the future.
Pictures three and four show how I used a some spare gussets to act as clamping surfaces to draw the beam in tight with the corner posts. Butt joints (e.g., screws, dedicated joint hardware) are good because they draw the pieces together. Gussets and mounting plates don't do that, so you have to add in the appropriate force to make a tight connection. This was about the only thing that went smoothly with this work step.
Next to last, you can see the alternate use for the 4590 gussets. My eBay special SR30 rails and carriages don't have a compatible set of holes for any of the Bosch/Misumi extrusion series I am using. However, the Bosch 4590 gussets look like they will act as a good interface between the carriages and the gantry. The offset load shouldn't be a problem (well within spec, even with a pessimistic load for the gantry), but we will see how that bears out over time.
Finally, the results of the acid bath. There was a lot of surface rust and some tar-like substance (baked-on grease?). The rail is working a lot better, but does have some surface pitting.
More in a few days.