I suggested wirewounds for durability.
Important to note that because this is an interaction of inductive and capacitive reactance, when the machine is running, voltage across the capacitor banks can swing higher than applied voltage. I haven't had any of mine swing past 2x across the capacitor banks, but it's theoretically possible, and when the contactor bangs open and closed, the collapsing field could sock it with spikes higher, so there's no reason not to be conservative in ratings. On my first CP-200 conversion, I WAG'd 15k 5-watt, with expectation that one soldered across each cap would survive and protect an individual against something bad, and as long as the interconnections were good, a failed cap in ONE place along the bank would be bled down by the others elsewhere in the bank.
The concern I also have... is that in this conversion, some machines switch the primary, some switch the secondary... if the primary is switched, the cap charges can hold long enough for a user to reclose when the phase of input AC is at some other potential than that of the capacitor, which causes a really hard 'bang' of current through the wiring. I have not yet experienced a failure on any of my machines, but that doesn't mean it won't happen.
One concern that I've always had, is the connectors on a capacitor, and how the capacitor 'banks' are interconnected. I HAVE found the push-on connectors to be variable in their reliability- most very good, but sometimes one connection goes bad. I've had some that were suspect, but apparently welded themselves on, and solved their own problem.
I WILL note that series resistance of the connectors IS somewhat of a concern also- when I wire up my cap banks, I'll daisy chain capacitors together, I'll put ONE connection lead at ONE end of the bank, and one at the OTHER, so that I don't wind up with the 'close' capacitors carrying all the load, and the 'far' ones not.
Keep in mind on these, that the bleeder resistors are ALWAYS carrying 'some' current... whenever there's power applied to the primary... and whenever there's a charge in the capacitors.
If just ONE is burning up, then there's a reason... could be as simple as a 'bad' one, or could be some other problem, but in order to burn a resistor, one has to exceed it's ability to either dissipate, or insulate. Nobody really pays much attention to insulation capacity of a resistor nowdays, because there's not much 'high voltage' flying around in our battery-powered gadgets, but realize that in a wound resistor, you have a coil, and that coil has turns which are 'spaced' apart, first, so that heat may dissipate, and second, so that voltage potential of one turn doesn't 'jump' to the next. If you put a high-voltage resistance tester (an HV supply with current limitation), you'll see that once a certain voltage is exceeded, the resistor's value drops rapidly, because the turns are shorting. This doesn't happen much with DC, but when you get to alternating current, and particularly high frequency (as low as khz, and certainly HF), you'll get current flow that bypasses the windings and just flows right by unresisted. I doubt this is happening, but if there's a marginal connection somewhere that's causing high-frequency oscillation across the cap, you may be seeing something like that. An AM radio playing nearby will give you a nice audible hint.
One thing that perplexes me... Why do your capacitors have a + sign marked on them?
