My thought was always to put film caps on a larger PSU board. CDE and AVX among others make such caps, readily available from Mouser, DigiKey, etc. If you dig into the data sheets, you see a significant tradeoff of lifetime vs. voltage rating, and it's different for Mylar vs. polypropylene. Mylar is more space-efficient, and might serve for the voltage doubler caps, keeping the polypropylene for the last caps, closest to the audio. I have not done the work to look at good choices of cap sizing - I'd prefer to get better filtering with the PSU choke, and not give up more capacitance than necessary. Making a good tradeoff here is a complicated subject.
For iron layout, I am strongly opposed to anything on the topside that has exposed terminals - they are at lethal voltages and should be kept out of reach! Enclosed transformers and chokes with wires exiting directly down through a hole are generally acceptable, though the UL would probably not approve them. Note that the upgrade iron, both transformer and plate choke, could be built this way - but you'd have to talk to Mike about that.
Putting the PSU choke topside in back, as johnsonad has done, is a little risky - the high voltage is not directly exposed, but the easily damaged winding and bobbin are not protected by a steel bell end. It would certainly not be acceptable to the Underwriter's Lab. From a performance perspective it seems to work well.
You are talking about a DC heater supply, not a cathode supply, by the way. :^) The 6.3v winding could be rectified and run through a chip regulator.
In theory this could power some time delay relays to delay the B+ and to short the output transformer until things have stabilized. However, I did try this a couple years ago, and the surge when the B+ came on caused all the relays to drop out and the timers to re-start. We didn't at the time have the leisure to figure out exactly what went wrong and how to fix it, so I'll just say there's a bunch of engineering yet to be done on that. I still think it's a good idea, and I'll continue to work on it. Incidentally, another relay could then be used to clear the soft start so that the 20 minute wait time before re-starting would not be necessary.
I have a few other things in mind, too untested to talk about yet.
