There will be a cumulative leakage current into the water from multiple boats even though each one has a leakage less than the 30 mA necessary to trip the GFPE in their pedestal. Apparently that's one reason for having a 100 mA feeder GFPE , as well as for leakage from a feeder itself into the water as was mentioned above.
With a 120/240V system, the leakage current to the water from one phase can effectively cancel leakage current to the water on the other phase within the current transformer of the.feeder's GFPE. And so using 2-pole 100 mA GFPE can allow even more than 100 mA of leakage on each phase through the water, as long as the difference in their currents is less than 100 mA. That is not really a good thing. Leakage currents in the water create a voltage gradient across the water. Fresh water is more of a problem than salt water because it typically has a higher resistivity than the human body does, and therefore a body will shunt a larger portion of the local current flow than in salt water.
A 120/208V 3-wire system would produce only a limited amount of leakage cancellation within the feeder GFPE even when the leakages on each phase are equal. Complete cancellaton would require no leakages into the water from line to the earth, but only leakages through the water between L1 and L2. That's not going to happen in a marina with a system that has a grounded neutral .