I suspect you have a classic case of the generator(s) having a third harmonic voltage content. You did not mention the ground connection of the damaged transformers and I will assume it was also solidly grounded. Is this true?
Let' examine some typical situations with round rotor and salient pole generators. You will find that for U.S. designed machines, the third harmonic voltage for round rotor machines can be from 2-5 % of rated line-neutral voltage at no load and this will increase as the load increases. The full load third harmonic voltage may be 4- 10 % of the the line -neutral voltage. Your hydo machines are no doubt of the salient type design and thus have the capability to have more third harmonic voltage. This is typical of salient pole machines. Remember that the lack of symmetry in the magnetic circuit will cause a harmonic content in the voltage waveform and these flux components will produce the fundamental frequency component and all of the odd order harmonics (3,5,7,9,11,etc.). There is a difference in machine design philosophy between the US machines and their European counterparts with the European units, in my experience, having a higher third harmonic voltage content.
Because the 5th, 7th, 11th and all non-triple harmonics will add to zero at the generator neutral, these harmonic voltages will not produce a circulating current in the system neutral.
In general, the third harmonic voltage of the generators is the only source of circulating current in the neutral of power systems.
What you need to do is to model the system by creating a diagram with the generator voltage source being the third harmonic voltage. The reactances in the circuit will be the generator zero sequence reactance multiplied by three to account that the frequency is three times normal system frequency and the zero sequence reactance of the transformer multiplied by three. The zero sequence reactance of a core type transformer is approximately 90 % of the positive sequence reactance. You mentioned that the transformer was a wye-wye design with both high and low windings solidly grounded. Does it have a delta stabilizing winding ( a delta tertiary)? You should be able to get the zero sequence impedance of the generators form the generator test/data sheet. Typically the zero sequence impedance is about 1/3 of the positive sequence impedance.
You mentioned that a full load the neutral current is 120 amps. When you calculate the effective current in the transformer it should be determined as the square root of the sum of the squares. For a transformer full load current of 875 amps, the effective current with a third harmonic current of 120 amps is 883 amps or about 1 % above nameplate value-hardly enough to cause the transformer to have any problems.
For the generator a harmonic current of 120 amps would cause an overload on the generator of approximately 6.5%
I am puzzled as to why someone would design a system with the generator and the interconnecting transformer both having a solidly connection to ground. This is rare indeed.
There is a way to limit the circulating current and that is to size and install a grounding reactor to replace the solid connection. The impedance of the grounding reactor would appear in the zero sequence circuit with a multiplier of nine time the fundamental frequency reactance. I assume you do not want to replace the transformer with one having a delta connection on the 3.3 Kv winding. Also I might mention that this is a perfect system to have a delta connected winding at the 3.3 KV side and a zig-zag grounding transformer connected to a resistor as the ground source at 3.3Kv.
In closing I also suggest that under no circumstances do you ground any more of the other generators. This could cause some serious problems with circulating currents betweeen the individual generators.
If you have the system data I would be happy to make the calculations for you.
I hope this helps