Teebee should probably chime in on this, as I believe he has tried printing at home.
In general, though, the more expensive a printer you buy, the higher quality the print, although there are other issues and exceptions to any rule.
If you are looking to mass print relatively chunky parts where accuracy can be "thereabouts", buying a printer could be worth it, although like ink printers, it is the material that will cost you in the end. If you are doing a lot of parts (figures, for example) where supports are needed, you could be looking at as much as 50% wastage in material for supports. Similarly you can expect a percentage of failed prints, especially during the period where you are getting used to how the system operates. There are numerous articles online describing add-on features that cost a fair amount to try to alleviate the failed print issues (heated build plates, coated build plates, special nozzles, free-flowing materials, etc.). When you factor in the cost of the materials, you might find the 3D printing services are cheaper.
There are also serious issues with the environment. Depending on the material you are printing, there can be serious outgassing during production...I have read of at least one case of a 3D printer killing the operator due to inadequate ventilation in the work room.
Accuracy is the other main issue. In 1/24 or larger you can get away with home printers. In smaller scales the minimum dimensions and accuracies are pushing the boundaries of believability. Using FUD, which is capable of dimensions smaller than any home printer, I recently worked on a wheelbarrow model. In 1/24 and 1/32 the handles could be a scale 1.5" in diameter. In 1/48, the handles were a slightly chunky 2" in diameter. In 1/64 the handles would be 2-5/8" diameter and in 1/87 they would be 3.5" in diameter. As well as this, the nature of the design is such that it would require significant support structures to print at home and would likely have a high failure rate as well as a high material usage rate. FUD is printed by stereolithography in support wax, which is not available in home systems yet, but which allows this fine detail and doesn't require permanent supports (the wax supports the parts during production and is melted away and reused after printing).
As much as the Form 3D system looks like the ideal solution, I am leery about trying it. The Shapeways equivalent has serious issues with the breakoff of support structures. I have written on the Gnatterbox about these issues before. The liquid acrylic printing medium is very brittle once cured.
One alternative I have considered is master printing via Shapeways. You could either get a part printed in FUD or in casting wax through Shapeways and use that as a master for casting, either through investment casting with the wax part, or mold casting with the FUD parts. Most major craftsman model railroad magazines have published articles on casting your own parts in urethane, plaster, low-melt metals and so on. The only difference is that the masters can be made via 3D printing instead of by hand in Styrene or wax. You could even do "part multiplication"...3D print one part as a master and make a singleton mold, then cast 10 or so and use them to make a 10-piece mold. There is even the option, although I haven't tried it yet, of making a 3D printed rubber mold. Receive the mold from Shapeways and then start mass-producing bricks, ties, and other scenic parts in plaster.