So I've been doing some experimentation and decided not to go with the previously shown active-clamping arrangement - or indeed anything involving a zener diode.
Firstly I tried using just a zener to clamp flyback to ground, but quickly realised that when PWM-ing the solenoid (to reduce the hold current), this makes the zener diode have to dissipate far too much power, due to the amount of current going through it all the time. Yeah, that zener diode got mighty toasty!

Maybe it's not so much of an issue at lower PWM frequencies, but in order to avoid audible buzzing from the solenoid, I was running the PWM at 25 kHz.
That then got me thinking about whether it was even worth trying the active clamping arrangement, as PWM-ing would surely increase the power dissipation in the MOSFET too. In fact, I suspect that a part such as the NUD3124 that uses active clamping is not actually suitable for PWM control of an inductive load. The datasheet doesn't even mention it. And the fact it has a 10K series resistor on the gate probably wouldn't help either (increases turn-on time?).
In the end I've just gone back to an anti-parallel diode across the solenoid, and that works very well with PWM.

What kind of diode best to use, though? I've never really given it much thought in the past when doing basic on/off relay control. I read that when you're PWM-ing an inductive load, the reverse recovery time of the diode should ideally be as small as possible, otherwise there can be detrimental side-effects. Don't quite understand why or what those effects would be, but heeding that advice I went to check some datasheets for the variety of diodes I had to hand. Okay, 1N4007 not so good (in the microseconds); 1N4148 great (nanoseconds), but possibly not stout enough; 1N5819, wait, what's this, no Trr spec on this Schottky diode?

Ohhh... they don't have a reverse recovery time.

I was already using one in the experiment, so I should probably keep it.
By the way, while doing some more reading into active clamping, I came across this circuit:

If I am interpreting it correctly, the high flyback voltage causes the zener to conduct, turning on the PNP transistor, which dumps the remainder of the transient through to ground. I suppose you would use it when you want the fast turn-off of a zener, but need a higher capacity.
But one thing is puzzling me: what is keeping the base of the PNP off the rest of the time?Edit: No, I'm a dumbass - I realise now. That resistor is acting as both a pull-up on the transistor base and a current-limiter for the zener.