Nominal 42V RMS could go over 46V during high line conditions and due to the transformer's regulation factor, even higher if lightly loaded. You therefore need a solution that can cope with 70V or even more after rectification. One approach would be to PWM the relay coil with a low side MOSFET, adjusting the duty cycle to average 24V across the coil. The PWM controller could be as simple as a NE555 running from a dropper resistor + shunt Zener supply, set to give a fixed duty cycle, or it could be much cleverer, measuring the rectified but unsmoothed 42V and adjusting the duty cycle continuously to maintain 24V average across the coil, removing the need for a large reservoir capacitor.
However the easiest way out of the hole you are in would be to see if you can get a solid state relay beefy enough to do the job, controlled by a dropper resistor from the 42V RMS signal or if not, use a small SSR to control coil power to a relay with a mains voltage AC coil.
I'm not too fond of Psi's ideas - half wave rectifying 42V RMS gives you a waveform that peaks at close to 60V, + a lousy power factor. In fact the RMS secondary current will be over three times the DC current, so you'd better be sure the 42V secondary is good for it. The 60V peak is more problematic. to get the average down to 24V +15% to keep a 24V relay happy, its essentially going to have to drop to zero between half cycles, so unless it has a slugged coil, that relay will be buzzing like an angry wasp, hammering itself to death. Also 4000uF is far too much - with 1A load current that's only going to give about 5V pk-pk of ripple, when fed from a half-wave rectified 50Hz supply. Note that the ripple current in the capacitor will be very high - comparable to the load current.
>70V rated buck converters are available, but certainly wont be cheap as a pre-built module. Sourcing the OEM relay may well be cheaper . . .
Edit: corrected a couple of numbers for 50Hz supply