[Chris is me]

In terms of viability of an exact center drive design, belts and chains can't be directly compared. 25 chain absolutely does stretch over time (and sprockets wear) and thus an exact center chain drive is not always viable. In a WCD, the small sprocket sizes use combined with the loads involved make exact center chain drive a bad idea.

Exact center belt drives are a lot more viable. Belts will not stretch in an FRC robot's lifespan. If you can machine with decent accuracy, you can hit the tolerances required. If I had to make up a number, I would say +/- .005", but really it's just never been a problem for my (former) team. We just CNC the drive tubes to exact center distances and it's good enough. Basically, if you have a CNC mill, there's no reason you can't do an exact center belt drive if you wanted to.

In fact I think it's easier to mess up tension with a sliding block belt drive than an exact center drive. Exact centers are probably better than the adjustment you can do by hand, and it's easy to over or undertension a belt. I think sliding tensioners for belt drives are almost strictly worse than exact centers in my experience. Counterintuitive, I know.

It is possible to overload a belt, causing ratcheting or belt failure. A rule of thumb is for 24T pulleys or smaller in a drivetrain, you will need 15mm wide belts. The combination of 24T pulleys and 15mm belts has served my (former) team well for several seasons, not once ratcheting, failing, or otherwise ever needing maintenance at all.

Other than for retention purposes I don't think the bearing holes have to be within .002" of perfect to work for exact center belt drives - that tolerance is probably a bit tighter than required. Still not hole saw tolerances though.