Ground

One of the possibilities is some unfortunate creature finding itself across phase and ground; there would have been a loud BANG at transmission voltages and it wasn't heard here.

Has anything been killed recently?

You have a ground fault alarm when the load is greater than 50MW? Did it happen once, or is it repeatable? How do you vary the load, by adding a dummy load? If so, could there be a problem with the dummy load?

You have not written how the ground fault [GF] relay works and whether it is limited to the generator alone [unit generator transfo isolates] or there is auxiliary power transfo direct to generator terminals?

Usually, generator ground fault relay detects displacement of generator star point from its theoretical zero voltage. Do you have grounding resistor or transformer - are they OK [expected current in resistor for voltage]?

Harmonic voltages do not balance out at star point & unequal fundamental voltages are possible - there is a limit to lowest GF voltage setting without risk of false tripping on external faults or unbalance [<5% of star-line volts difficult]. GF relay usually has harmonic filter and time delay.

1/ Have you checked that GF relay actually works at scheduled volts & time delay without frequency effect?? Have operate points changed since last check?

2/ Have you put another meter or recorder in parallel with GF relay to see changes with time of day or load which show "creep" towards GF settings? Are there any previous GF voltage records to compare with present?

3/ Or an oscilloscope to view GF input waveform/harmonics change with load? How compares with any previous record?

I would guess that the ground current or ground voltage is sensed, and that it is marginal below 50MW and above the limit above 50MW. Check your guidelines for measuring resistance between ground rods. Use equipment designed for this purpose (not a VOM).

Barring some malfunction of the ground sensing relay itself, the direct correlation between ground fault and power loading (use amperes line current rather than MW or kW) is the dimensions of your conductors increase with the higher loading (expand with temperature). For instance, if some part of your cooling is blocked, one or more of your stator windings could be moving into a position uncomfortably close to your grounded stator laminations, perhaps some of the insulation at one spot has started to carbonize, and the growth of the conductor has exceeded the design dimensions.

The risk in allowing this to continue is allowing the 1 phase to ground to escalate in to a 2 phase to ground, which will then take your generator out of commission and cost significantly more money to repair.

Improperly constructed or improperly installed connections,stress cone connectors,or dirty insulators.

Routine scheduled cleaning and inspection and torque check of bus bars is recommended,as well as megohm testing.

Could be a small nick or defect in the cables or terminators.The more amps,the more capacitive effect on the insulation.The semiconductor portion of the connectors are designed to bleed away the static and to prevent corona discharge,which would ionize the air between conductors and create a major fault.

The leakage detected may be a warning of a future catastrophic failure.

I would recommend immediate resolution of the problem.

The recommendations for a thermal scan are valid and probably a quick way to gather information,but further action will be required so schedule some downtime in the near future.