Think of a single phase transformer feeding an open circuit on the secondary. The AC is a normal AC signal then suddenly the AC signal instantly becomes 0 V (this represents the short on the primary feed). There will be some amount of magnetic flux (aka energy) in the transformer core that will push a current back out of the primary windings into the 0 V AC source.
If the transformer is perfect in all respects (other than it has a magnetizing inductance), the current will flow into the "0 V" indefinitely BUT there are of course resistive losses so this current will be an exponential decay to zero. If there is no leakage inductance on the primary winding, the 0 V (aka short) will ensure that there is nothing seen on the secondary other than the cessation of the AC waveform when it instantly falls to zero.
If there is leakage inductance on the primary (normal of course) then there will be a small kick-back voltage seen on the secondary due to the magnetizing inductance of the primary not being perfectly shorted to 0 V.
With or without a load it won't make a difference - there will be a small kick-back voltage seen on the secondary as the magnetic flux (and current) exponentially decay to zero.
It should also be noted that if the AC falls to zero volts at the very peak of its waveform, the current in the magnetizing inductance of the primary is zero and no effect will be seen. This is because, at that instant in time, the flux in the core will also be zero. If, on the other hand, the AC voltage halts at a zero-cross (and stays at zero), the flux in the primary will be at a maximum and the effect described above will take place. At all other points the effect will be proportionally less.
Would highly recommend that you try testing first with 2.4vac or 24VAC input. For a simple unloaded transformer the ratio of input to output will be the same. Note that determining proper 3 phase output may require a scope or special circuitry to see the phase differences (if needed). Using a Variac (variable transformer) can help select low input voltages. Other safety measures would include fuses and/or resistors in line with the input voltage.
If absolutely required to have a test system using the full voltage input you might put the whole transformer that is being tested behind a clear shield or enclosure that prevents any test person from accidental contact. Entry to the enclosure or removal of the shield would open interlock switches (opening up all input lines) disconnecting all power to the transformer.
Best Answer
The cause could be a broken neutral wire shorting with a line.