There is an issue with run-to-run consistency that you should be aware of.
For WSF in particular, there are two different models of printers, one that can print LARGE items, and another one that is more limited in size.
Unfortunately, the tolerances between the two printer models are different, and you can't be guranteed which particular printer your item will be printed on for any one order, or even that two items on the same order will be printed on the same printer at once. Just because it's a smaller item won't necessarily mean that it's always printed on the SMALL printer. Therefore a hyper-exacting repeatablity won't be possible.
It is also highly likely that the printers have to be regularly recalibrated, and that the accuracy for even a single printer could drift slightly over time.
Someday, we'll have sub-micron accuracy on parts 1000 meters long. The technology just isn't developed to that level yet.
On many of the material pages is listed "Accuracy". I'd go with that for now as the allowable run-to-run differences.
But, it sounds to me like a request for "high definition WSF" or the like would be a good idea. I'm sure that will be added to the list (at extra cost), but it'll be somewhere lower on the priority list than the request for user-selected print orientation (which, by the way, also affects accuracy).
I also doubt that Shapeways has incurred the expense of measuring a statistical sample of printed items such as to be able to satisify your "dart" example. These are "professional" machines, not the home-grown devices you may have seen elsewhere. Shapeways depends in large part upon the specifications and accuracy provided by the manufacturers.