Electronic – To what extent to I need to be concerned about PCB punch-through when dealing with AC mains

There are various standards. Some companies I have worked for have claimed to go thru them and determined that 5mm spacing for "line voltage" was going to cover all the cases. That is deliberately a superset, so of course there will be individual standards less tight than that.

If your design can tolerate 5mm space, then do it and you're done. If not, you have to spend some time figuring out which standards apply, and then check them carefully to see what the minimum required spacing is for your particular type of device and target usage. For example, home and industrial often have different limits.

There are several reasons spacing requirements might be larger than you would think. The different levels of importance placed on these has something to do with different specifications.

1. Line voltage spikes. You say your line voltage is 220V, but that is nominal. At least 10% should be expected, which brings it to 242V RMS. The peaks of that are +-342, and that's with everything working normally. In reality power lines can have fairly large spikes on occasion. How large depends on what you believe and what probability you want to protect against.

2. Rectification. From above we have to expect +-342V on a good day. However, some line circuits may couple things such that one peak or the other is held at ground, so the worst case the high voltage may be double the ground-referenced line peak, or 684V in your case. Whether this is considered, allowed, or made a separate catagory varies.

3. Dirt. A perfectly clean PC board has quite high resistance and therefore would support very little leakage accross even a small gap. However, dirt is a fact of life, and accumulated dirt will significantly lower the insulation resistance, especially with high humidity. You may think that 10 MΩ/square is pretty good, but if two such traces are running next to each other for 10x their spacing, then you've got 10 squares in parallel and you're down to 1 MΩ. Again, how this is considered in a spec varies. How dirty is dirty? What are the consequences of a few 100 µA leakage? The answers differ, and thereby so do the specs.

Your wattmeter only has arc suppression gap set at 6kV standard. So you can expect anything less than that and possibly more , likely in Florida , with transients that do not ionize the gap.

This is why burnt out LEDs in Florida are so common with cheap PAR Lamps.

One way to suppress peak voltage is via line filter using two X1 caps and a choke or a 3rd order LPF. This reduces the 10kV spikes to <3kV or even as low as 300V but stretches them out in time.

So Optoisolator products assume you know about this and are all rated for 3kV isolation.

Any protection you add to clamp these voltages must be able to,handle the current that follows.

Gas tubes can handle 10kA surges but expect you to fuse the input.

MOV's are cheap and good but can only absorb occasional surges and will fail after so many accumulative Joules in energy. Chokes allow the rise rate in current to be limited until they saturate , then they behave like wirewound resistors.

Here is some industrial design background for line filters. Line filters in LED bulbs must pass these tests, and not all,suppliers are legit in saying they are compliant.

There is no "one solution fits all" because of other requirements for low cost, high reliability, and new EMC compliance rules for devices >= 100W must have power factor >=0.9 using active PFC etc.