Upload & Sell: On
There are at least two schools of thought on the matter. One, of course, as you know, is to use the widest gamut possible to "preserve" everything that was in front of your camera. The other is to use a gamut that is more conservative but still provides for great looking, saturated prints and more or less matches the gamut of your printing devices.
The people that subscribe to the first scenario are often paranoid about losing any bit of detail, and many times make ProPhoto files that will never be printable on any device in our limited lifetimes. So, it's great to have a wide range of colors, but if you can't ever reproduce them, is there really a point? Secondly, there are often issues when moving from a super saturated ProPhoto image to a file that's going to print on a limited gamut printer like a Fuji or Lightjet. When the out of gamut areas are too far out of gamut, that's when you can actually see the effects of the choice of rendering intent on the output. You either clip off saturated detail (RelCol) and end up with areas of flat color, or you compress the gamut (Perceptual) and compromise on the accuracy of your in gamut colors, which have to be moved closer together to maintain color separation. And if you take a super saturated ProPhoto and convert it to any other working space made with a matrix profile, you're only going to get a Relative Colorimetric conversion anyway.
Rather than choose a blanket, one size fits all approach, it's often more effective to choose your working space based on the type of image you're making. The plain fact is, is that most images just don't have the gamut range that requires a super large gamut RGB working space, and even when there are small areas in an image that do fall outside of something like sRGB or Adobe RGB, they simply have no visible effect on the output.
About a year and a half ago, I made a series of large prints that I sent to another FM member when we were having a discussion regarding the utter horribleness of sending jpegs to print. Not only did I make a series of prints from progressively lower quality jpegs, but also comparing sRGB, AdobeRGB and ProPhoto on some saturated images. The conclusions were surprising, both to me and to Phil. It actually took getting down to something a jpeg level 4 quality to actually see anything in print, and the differences between sRGB and ProPhoto on a large, saturated image, were so small that if you didn't know what you were looking for, you probably wouldn't have seen it. And, if you weren't printing on a large gamut medium - in this case Lexjet eSatin on an Epson 9900 - the differences would have been even smaller.
My suggestion is to make a series of test outputs of your typical types of images and see if you can see a difference, and if so, how much. You might be surprised at how much hype there is out there, especially from those purporting to be professionals.