I'm a bit confused how to make a snubber for the switch. Do you think just a 22nF cap would do the job? No resistor, nothing else? People say resistor needed to reduce peak current when switch is closing. Alternative suggestions are to put cap in parallel with trans, but I have doubts, as it becomes a resonant circuit, imho. So, some calculation is required. I heard that cap can only be 1-2nF as it becomes a capacitor divider with the switch. Since switch has only few pF, it supposed to be enough. But I didn't really see any proof that this gonna work.
As for other concerns. Why I have slots is because I may need active cooling, fan may suck quite a bit of dust (but I'm trying to put heatsink with the fan outside of enclosure). I also want to make a "robust" design that doesn't make assumptions, just an "unconditionally save". It is 2.5mm wide because even if the board gets fried by a lightning and there is a deposit of conductive carbon, there is still enough clearance. I'm not sure if this reasoning is correct, let me know if I'm wrong. I just want to have one softstart circuit that I could use in all my devices without much thinking. Trapping flux is not a problem as I do hand-soldering and washing. So, for a production unnecessary slots can be bad, but so far I'm only gonna make two of these boards. I already spend two weeks for it, so a few more hours for assembly is not a problem.
Slots indeed weaken the board. I hoped that relay and transformer would reduce board stress a little bit. Now I'm not so sure that stressing them is good idea.
I moved spade connectors away from mounting holes. Even though I use nylon spacers and screws to mount boards, someone else may use this board. So, thanks for advice!
The small transformer that provides isolated floating supply (no connection to ground) is inherently short-circuit proof. It doesn't mean it cannot fail. I used to have a 0.1A PTC, but removed it for the sake of simplicity. I think even if there is short inside, mains will either melt wiring inside, or fuses will blow.
Why delay is 0.1s. It's five mains cycle, simulation showed it should be enough to avoid the biggest portion of inrush. I still changed values of time-setting resistor and capacitor (it's 1uF cap and 100k resistor instead of 100nF cap and 1MEG res) as 1MEG resistor seems to me a bit too much and may be susceptible to noise? Or may be not as 100nF was still quite a lot to be quickly charged by emi?