I agree, but this is an unusual situation where there is little or no voltage overhead in the secondary circuit and during the beginning of the charge cycle the transformer will be under almost short circuit conditions. The optimal solution has already been suggested, a voltage monitor across the capacitor that controls a current limiter in the transformer primary. I suppose the simplest solution is a thermistor on a heatsink to slow it's temperature rise but my main concern would be the transformer burning out while the current is so high. When the capacitor has charged, the primary current would be very small so even if the thermistor resistance increased again to say 25 to 50 Ohms, it would only drop a small voltage.
A slow reacting inrush NTC would be the ideal component but as far as I know none are available. I don't consider relays shorting out milliohm resistors in the secondary a good solution, the relay contacts themselves may have higher resistance than the resistors they shunted. I would also rule out any active power control in the primary because it would either dissipate a lot of heat or be pulsing current into a very low impedance inductive load. Whichever way you aproach the problem there is no 'simple' solution but many complicated ones, at least in consideration that all this does is charge a capacitor.
Brian.