Let's break this up into separate issues.
If the box contains splices, and the box is metal (where the raceway or cable is not), then the ground needs to be tied to the box. Makes sense; if another splice comes loose, you want it to trip the breaker, not keep the box energized.
If you're using pipe, or AC, then you need not do that, as the raceway is the ground path. In this example, you're NOT tying all the ground paths together in the box.
If you're having another circuit pass through the box, without splices ... I see no need to tie that circuits' ground wire to the box. You've already got that box bonded. Whether you have transformers, generators, multiple service drops, whatever ... the grounds of all the systems are supposed to be tied together at the sources. If they're not ... I'm not sure you want to do that at some remote, unknown box, using tiny wire and a little wire nut.
Now, let's assume that you have two separate systems in the box, both with splices. One word comes to mind: PARTITION. Isn't that box supposed to be partitioned? Even when the one system is 277, and the other 120, it's possible for there to be more than 300v. difference between the two (phase differences). With the box partitioned, I can see where you would make separate bonds to the box.
Why have more than one wire nut tying the ground wires together? Well, wire nuts are limited in their capacity, and it can be a challenge making a good connection of five or six wires. You've got to start thinking about box fill at this point.
Scott's right to emphasize the distinction between groundingING (green) and groundED (white) wires. The white wires ought to be kept separate in the box, joining only at the panel. While the code might allow some wiggle room on this point, you're introducing an opportunity for all manner of bad things to happen.
