Feeling just slightly cocky, this evening. 
Printed an updated “boat” for my next-printer. Had to angle the print for better strength and more accuracy (with a dubious PLA filament). It worked.
Also, the print took about eight hours.
Getting more practiced at re-stringing the belts. Helps when your boat is not threatening to crack apart. Got decent tension on the belts, by hand(!).
(There are a few bugs in my prints. They will be corrected.)
The end result moves well in X and Y, with no play. The level of friction is high, but if I just wanted a 50mm/s printer, this might be sufficient.
I want a lot more than 50mm/s, but if result is this good for plain PLA structure, should do a lot better with better bearing materials.
Need to sort out the Z platform next. After that, will have no choice but to sort the electronics and software.
did you look at the oil free bushings?
My father - who worked on borderline-impossible aerospace (military) problems before he retired, recommended Delrin as a suitable plastic. He ordered and received Delrin rod, and has more than sufficient tools to shape.
So … after I have sorted the cruder issues, we will have a chat about machined Delrin inserts. 
I might also make a try at printed inserts using materials from Igus, as recommended from the community.
Delrin (or POM) would do very well, I think. But I was looking at your tensioner, and, although I like the way you think, I doubt it would work very well.
This is the solution I came up with:
missing/deleted image from Google+
Torsional nonsense is fine as long as it’s small. you’re never going to get the center of mass, center of stiffness, and center of friction all exactly coincident. You should try to get close (which you seem to be doing well) but small torques are inevitable and just not a big deal.
Delrin is a great bearing material because it’s long-wearing and has a lower coefficient of static friction than dynamic friction, which means it is more or less immune to stick-slip behavior.
Printed nylon or printed Igus tribofilament works really well too.
@Rien_Stouten While your belt-tensioner is undoubtedly better than mine, my inclination is to build a jig to drop the belts into the printer under tension. In effect, the printer is simpler as I expect to print a tool (jig) to apply tension to the belt.
As yet, I have not seen guidance as to the appropriate level of belt tension.
@Preston_Bannister For rigorous purposes, the proper belt preload tension is slightly higher than the net tension applied by the motor at full stall torque. This guarantees the teeth can’t jump under any achievable combination of position and load. But nobody ever bothers measuring 3d printer belt tension – just get high-quality belts nice and tight and they perform fine.
@Ryan_Carlyle My first guess was to expect a resting belt tension similar to the added tension of the motors at full torque.
Then I remembered that belts in this category are meant to be remarkably inelastic. If an un-tensioned belt is of same dimension … is there any way that can be a problem?
Of course, my current no-spec cheap-Chinese belts are probably reinforced with dust-bunnies, so not expecting much there.
@Preston_Bannister Cheap belts are stretchy, and your printed plastic frame / motor mount is stretchy. In tandem, yes, you can get enough stretch to cause tooth jumping. But in practice tighter belts mostly just improve print quality by eliminating tooth backlash and getting the loads fully transferred from the soft elastomer teeth to the inelastic tensile member in the belt.
@Ryan_Carlyle My expectation is that when near final form, I will order a proper Kevlar GT2 belt. (Yes, Kevlar.)
And I will print a jig to pre-tension the belts.
@Preston_Bannister Kevlar is fine, but if you want stiffness, genuine Gates GT3 is about 10-15% stiffer than GT2, and 9mm wide is about 50-60% stiffer than 6mm wide. (Yes, more than 50%, because the spiral tensile member lay causes more de-rating the narrower the belt is and the fewer complete wraps are spiraled.)
