John,
This is a good topic to explore. I have done most of what you have suggested several times. I have had a couple of issues that I will present. I typically use 1N4007 diodes because I have a pile of them. Being 1 amp @ 1KV rated they can work in a lot of places. The ones made since the early 1980's use ion implant technology in the manufacturing process (was diffusion before) and these have a "controlled avalanche" break down characteristic. The old ones, once broken over from a high PRV spike usually die.
So these days we usually dispense with the resistors and capacitors across each diode for the reasons mentioned above. Still figuring the PIV needed based upon the formulas is the minimum. Doubling the number of diodes from there is good insurance. But what about that choke input filter spike when that big 8 Henry choke field collapses? The path has to go through those diodes in the reverse direction....Your point about adding a capacitor at the input side of a choke input filter to ground, like a .1 uf cap is a good one. This forms a series R-C snubber where the 'R' is the choke DC resistance. The time constant of that R-C could be adjusted by varying the "C' value.
Then consider those 866 Solid state replacements that are plug in. These have a sh_t load of diodes all in series, and they have a 'controlled avalanche" characteristic such that they can just take a lot of abuse without failure. They are old technology now, but they worked pretty well. Many a ham has made a stack of diodes that failed, when those SS replacements survived.
So back to your choke input conversion after Solid stating the rectifier. What has tripped me up is the output AC ripple is higher then before, so the output capacitor value needs to go way up, maybe 5X before or something like that. This is no big deal though. Maybe the output capacitor was 20 uf before, and going to 100 uf is cheap and easy with caps available today.
Another issue is the diodes each side overlap in conduction a little bit each half cycle. This might only be for 100 us, and sometimes approaching 1 ms depending on the diodes used. During the conduction overlap the transformer secondary is shorted out. When the reversed biased diode(s) finally stop conducting, the transformer generates a back EMF spike of high magnitude and short duration. This happens just after every zero crossing, or every 8.33 ms for 60 hz power. The result is a pulse train at a 120 hz rate that will go un-attenuated through the choke (the turn to turn capacitance allows narrow pulse passage), and the series inductance of the filter cap does little to stop the pulses either. So as a consequence, you might hear a "buzz" in the low level audio, or rf stages that was not there before.
So to cure this buzz, a series R-C across the transformer secondary often snubbs this out. Another method is to follow the L-C choke input filter with an R-C, making a L-C-R-C filter. The 'R' only needs to be 100 ohms or so, and the 'C' added can be a .1uf polypropylene across an electrolytic like maybe a 20 uf...
Edit: One other thing I did was to resonant the filter choke; this was on a dual supply like the NCX3 thread. In my case the LV choke was a 4 Henry smoothing choke, and with a .47 uf capacitor it resonated at just under 120 hz. The AC ripple went way down, and even the diode switching trash was way down. The problem here is that to keep the resonance rise down (voltage across the capacitor), you need some loss in the circuit. The DC resistance of the filter choke is effectively part of the L-C-R resonant filter where the 'R' value will lower the resonant Q. So adding some 'R' might be useful if the choke is low on DC resistance. I have heard that when doing this you want to be on one side of resonance. Perhaps if it were the high frequency side, then as the load current increases, the resonance will move further away since the choke inductance will drop. So maybe the low side is better, established at the minimum load current?
So with resonating the choke, the capacitor needs to be capable of passing high current, and tolerating a good bit of voltage. Surely an oil cap would do this. But these days we have AC line voltage rated capacitors at the input side of switching power supplies, and many are rated at 275VAC. These are polypropylene, and can withstand at least 2X that rating for DC. In fact I was looking one up a while back, and IIRC was rated to withstand something like 2KVDC for a short period.
In my case I did not heed my own advice, and I used a .47 uf / 400V Black Cat. It worked great for about a year, and then one day I had way too much voltage...The cap shorted.

Jim
WD5JKO