Power supply Inrush protection
Let's work our way left to right through this power supply and see why certain components are used and what characteristics are important in those components.
Step Start or Inrush Limit
R7 and RLY1 make the simplest possible inrush limiter. In this system RLY1 generally has a pull-in threshold of around 70% of rated voltage. This means all primary current flows through R7 until the capacitor bank reaches about 70% of full voltage. This greatly reduces inrush. R7 should ideally be sized to limit current to maximum safe operating current, or some amount less at your preference. If the line is 250 volts (there is no 220 in the USA, that went away in the 1950's) and the primary system safely tolerates 15 amps at worse case load, we would want R7 to be at least 250/15 = 16.7 ohms. Use the next highest standard value, or more if you prefer. If you too much resistance the amp will not reliably close RLY1.
Let's say we use 20 ohms with 240 volts. On switch closure when the primary looks like a near-short, inrush would be 12 amperes. If the supply was loaded with more than 1000 watts load before RLY1 closes, the relay may not close. Because of this it is a good idea to use an energy storage or pulse rated resistor that handles significant surges without failure, and to fuse the resistor (fuse is NOT shown) with a modest current fuse. Typically a 2-amp slow blow fuse will handle a 10-amp inrush system for the short duration of R7 current. Size the fuse as small as possible consistent with the fuse not opening on normal starts. The system below is a 120V system.
Whatever you do in a high-power variable-load capacitor input supply, never put the inrush protection or limiter as a fixed resistor in the primary or secondary!
A capacitor input supply operates with very high apparent power factor. Any resistance pre-filter capacitor will dissipate far more heat than most people expect, and will also cause a much larger voltage drop than expected.
A capacitor input supply conducts only over a small fraction of each cycle. With a short conduction angle, peak currents are very high. NEVER just add a series R on the AC or unfiltered DC side of a supply that has variable load, or has a high power drain. Adding a resistor that is in-circuit all the time on the AC or unfiltered DC side ruins transformer regulation while making excess heat. If you are going to spend money on a resistor suitable for the B+ line, put it on the DC output to limit sure (Ohm's law can be directly applied without consideration of power factor there), and put a real step start in the primary. Link to power factor heating of transformers and resistors.
Everyone is probably going to pick their own filter caps, bleeders, and diodes, often by gut feeling rather than science, so I'll not specify capacitors or rectifier design. There are many Internet suggestions that sound logical but are really are foolish, so be careful.
My only warning will be to NEVER use carbon resistors for equalizing resistors. Carbon is a semiconductor, and semiconductors have less resistance if subjected to thermal overloads. High dissipation carbon resistors can actually fail shorted. That's a bad thing on a power supply.
Note D2 and C2. Both are critical safety components. When properly sized D2 and C2 will protect the meters, meter shunts, and the operator under any condition of HV fault. Use the lowest voltage C2 available in a small disc capacitor. I typically use 50 volts. The value generally is a .1 uF. This capacitor serves two functions, bypassing D2 for RF and providing an additional clamp at 100 volts or so if D2 should ever open. D2 can be any diode that can survive, without destroying the case and opening, a current of HV over R4 plus worse-case path fault resistance. Fault path resistance would generally be the sum of filter capacitor ESR and R4's resistance.
R4 typically should be chosen to limit fault current to non-destructive values, typically in small amplifiers (below 5 kilowatts output) this winds up being a path resistance about 5 ohms per 1000 volts. A 4000 volt supply would typically require around 20 ohms fault path resistance. Typically that path resistance would be comprised of 2.5 ohms ESR in the capacitor bank and 2.5 ohms total in wiring, RF chokes, and other resistors. In this case R4 would be 15 ohms, which adds to the other wiring and component resistances for a total fault path resistance of 20 ohms.
With 4000 volts we would have 4000/20 = 200 amperes fault current. A 1N4007 would only handle 30 amperes before becoming unreliable and shorting, but that does not mean it would burn open. We really have to test diodes to see how much pulse current blows the case open. Better quality 1N4007's will handle 100 amperes for 30-50 milliseconds, and 1N5408's over 200 amperes for the same time period. If you are unsure, use a 1N540X series diode, like a 1N5408.
D2 may require more than one diode in series. If the grid or plate shunt drops more than 0.5 volts, you would want to add more diodes in series at D2. Allow 0.5 volts per diode.