As a matter of fact, we on 188 tried this very thing last year. It certainly looked good on paper, but we ran into serious issues with the gears grinding, and the centre distance of the gears being extremely susceptible to changing. One moment the gears would be in mesh, the next, the forces acting on the gear (due to rotation) would push the gears apart, causing a loss of contact. To combat this, we tried pushing harder on the gears, so that the gears would be "happier" while running. The problem with that was that the gearbox was now too tight to accelerate properly--the gears would just jam. (Then all of a sudden, as we tried to carefully adjust the spacing, they'd move out of alignment, putting us into neutral once more. At least the motors weren't getting damaged in neutral

.)
With regard to the jamming issues--don't even think of overpowering it, and then settling down to a happy equilibrium once the robot gets up to speed. First of all, it's a waste of power when you need it the most (i.e. acceleration from a standing start, or in a pushing match), and second of all, our robot last year had 6-motor drive (Bosch, F-P and CIM), and still couldn't overcome the jamming issues with much finesse. It comes down to the fact that your centre distance has to be within a few thousandths of an inch of the optimum spacing (quoted centre distance plus 0.003" to 0.008" is a good bet for the optimum value).
If you do want to try it, keep your tolerances very, very tight (there's no room for error here). Also, we used 20-pitch steel and cast-iron gears--you might want to try these, since despite the unusual forces exerted on the gears, they held up admirably.
Lastly, design the gearbox so that it can be locked into one gear (high or low) if all hell breaks loose. We made some significant last-minute changes to remove the shifter stage and set it into high gear, when we decided that it was too much trouble. It was good enough to come second at the Cdn. Regional and win W. Mich., even without a low gear...