For the benefit of others: This addresses a particular question posed, based on the OP's determination that having a charged capacitor is something that he wants to avoid and because of the observed behaviour. It's not intended as a 5 page debate over bleed-resistors-or-not.

"what size of resistor would be appropriate"

You need to solve a couple of equations...

Tau = RC

(R is the resistor value in Ohms,
C is the capacitance, in Farads
Tau is the value of time taken for the voltage in the capacitor to fall by 63% (or to 37%) of the original voltage).

So, for example a 20,000uF (0.02F) capacitor, and a 1k resistor (1000R), Tau = 1000 * 0.02 = 20 seconds to discharge a 100V to 37V. Then another 20 seconds to discharge to 14V (37% of 37V) and so on. The lower the value of resistor, the quicker the rate of discharge.

However, the lower the value of resistor, the higher the power dissipation that resistor endures whilst powered on, until the capacitor is discharged.

P = V^2 / R

(where V is the voltage across the capacitor)

Assuming your 24V supply.

P = ( 24 * 24 ) / 1000 (using the 1k resistor above).
= 0.576 Watts.

Therefore you'd need a resistor rated at above 0.576 watts - there are a 0.6W rating, then you typically find 1W, 2W, 3W, 5W etc...

Also consider then that you're dumping that power into your PSU enclosure as heat whilst the PSU is operating. To be honest, a couple of Watts is 2/3'rds of cock-all, but you need to be aware.

Myself - I have 72V DC power and, I think 10,000uF capacitors (x2, dual secondary/supply) - I really can't remember but expect I've used in the order of 2k2 wirewound 3W resistors. But that's just my choice/wet finger.


EDIT:

AHA, Neale - you posted first. Part of my rationale is that my cabinet is modular in design - as you say, if I isolate the drivers whilst I'm working in the cabinet, it's reassuring vs spot-welding the wedding ring to chassis.