I'm not going to pretend to be an expert, but I think a lot of makers out there want to design for high voltages but are afraid to. Others don't give the topic the respect that it deserves. I'd like to believe I'm somewhere in the middle. And for what it's worth, I'd like to offer my own perspectives on how to make PCBs that will handle high power safely.
First, we need to agree on what constitutes a high voltage. Once upon a time, I was doing some telephone wiring (this was back in the old days of analog POTS lines. Some of you youngsters will never have experienced a phone that had a cord that did something besides charge the battery). I had my fingers on the wires at the exact moment the line started ringing and I got a nasty buzz from it. Someone smarter than I was would have taken the phone off-hook before working on it. The take-away is twofold: high voltages are lower than you might think. Even without the 90 VAC RMS 10 Hz ring signal, POTS lines when on-hook were 48 VDC, which is just under what is often used as the 50 volt (peak) threshold for "high voltage." The second take-away is that you need take into consideration every possible state the line could be in. The on-hook voltage of a POTS line may be low enough to work with, but watch out when it rings! So for the purposes here, I'm going to take as my delineation a 50 VDC peak threshold. If at any time a signal is anticipated to have a voltage higher than that, then it's high voltage.
High voltages are hazardous because very small currents can be very powerful. The guiding principle, therefore, is to insure that any high voltage conductor is well isolated from anything else. If it's a wire in air, that means that its insulation must be rated to withstand the working voltage of the wire inside. For a PCB trace, that means that it must be separated from any other trace by enough space such that neither the air, nor any foreign matter that might be on the surface of the board, might cause an arc. In PCB design terms, this is called the creepage distance, and it is the distance between two adjacent conductors measured along the surface of an insulating material between them. A related term is clearance distance, which similarly is the distance between two conductors measured straight-line through air. Clearance applies, for example, to the uninsulated portion of two wires that are entering two adjacent pins of a screw terminal. Creepage applies two two traces on the board.
At this point, It's important to interject an important note. Soldermask is not an effective insulator. Soldermask's purpose is to prevent solder from bridging two adjacent traces and to protect the bare copper from corrosion from flux residue or the like. It's not there to impact creepage or clearance distances. For one thing, Soldermask is easily abraded or scraped away, and you obviously can't count on it having any insulating properties if it's not there. For HV circuit design, conservative engineering dictates that you pretend for the purposes of creepage and clearance distance measurements that the soldermask just simply isn't there.
By contrast, PCB material is a very effective insulator. Creepage distances do not count when traces merely pass over each other on opposite sides of the board. Effectively utilizing both sides of the board is an excellent way to help you achieve proper creepage and clearance. However, if you do put HV traces on the bottom of the board, then you must pay close attention to how you intend to mount the board in its enclosure once you're finished. You must insure that proper clearance is maintained with the bottom of the board. If you have no HV traces on the bottom, then in principle, you would not need to worry.
When designing for HV, proper design always begins with the schematic. Let's take a look, for example, at the HV Contactor board for OpenEVSE II:
Note the broken grey line that forms an area on the left side of the schematic. That area is the designated "high voltage" portion of the circuit. Think of the HV part of your circuit and the rest as being in two different countries. That line represents the border. And nothing is allowed to cross that border at all except for special isolated parts. In this design, there are 7 points where the border is crossed. One is an isolated AC/DC power supply module. One is an opto-isolated triac driver, four of them are standard opto-isolators, and the last is a ground connection that requires separate discussion. In all of those cases (except the special one), the components offer galvanic isolation between the two separate sections. This isolation is a basic safety requirement. The HV wiring is effectively a contaminant that you must keep completely boxed up and away from any possible human contact. It's only via galvanically isolating components that any indirect contact can be allowed. If you follow this basic rule, then in principle, the rest of the circuit can be deemed as safe and need not have special safety precautions taken.
There is one exception in this circuit to the "border" principle, and that is that this circuit includes a ground impedance monitoring function. A small, carefully limited current is allowed to leak from the HV supply to ground, and is measured on its way there. Normally such a situation would be an error, but in this case, it's a carefully constructed exception case to the rules of HV isolation. Exceptions to any rule can sometimes be made when other considerations demand them. They just must obviously be done with great care and thought.
Now, let's look at the board:
See that white line about 3/4 of the way up from the bottom? That line is the exact analog of the grey dashed line on the schematic. It's the border between the "safe" and "HV" sections of the circuit. Note that in the safe section, there's a floated ground plane, but there is none in the HV portion. This is simply because ground planes are an anathema to maintaining creepage distances. HV wiring should generally be kept clean and simple, and I can't imagine a circumstance where floating a ground plane through HV would be necessary or desirable. Note also that the only parts that are allowed to "straddle" that border are the galvanically isolating ones we pointed out in the schematic (the four opto-isolators are, in this case, actually four modules in a single component - the gull-wing SMD DIP-16 in the center. Note also, that it's actually drawn on the silkscreen as two duals rather than one quad, but it still works). Also, note that those parts don't straddle the line evenly. They're off-centered in a way that shoves the HV wiring as far away from the border as practicable. This is good for creepage to the ground plane.
OpenEVSE II actually consists of two boards. The second board is a logic/display board. In my design, one of the goals was to separate the HV and logic as much as possible. You can see the result of that in the design of the HV board above. There's as little circuitry as possible on the HV board, with the hopes that the design will need little, if any, troubleshooting. This is another hallmark of good HV design - KISS. Any time you need to poke at an HV circuit while it's powered up you're taking your life in your own hands. The less of that you have to do, the better. You can poke at the logic board all you want. In fact, because it's a separate board, you can have it sitting on the bench outside of the closed chassis that houses the HV board, which makes live testing reasonably safe.
Another good idea for design is to try and place some sort of safety device immediately inside of the ingress of HV, if possible. In this case, the AC in lines each terminate immediately in fuses before hitting the AC/DC converter. Additionally, the hot line hits another fuse before going to the contactor triac circuitry. The hot line of the relay test and the AC in also hit two large 150K resistors. Those resistors are spec'd as flame-proof resistors, meaning that they act somewhat like fuses when stressed and burn themselves to open circuits to protect the rest of the system. Note that, in principle, on the far side of the resistors, the voltage will be much lower, so we can get away with somewhat reduced creepage distances in the area around the opto-isolator pins, but having a respectful creepage distance to the ground plane section of the board is still a good idea.
The other part of high power is high current. High current is relatively easy to deal with. The guiding principle is simply P = I^2 * R. In this case, the power of interest is not the power of the current flowing through the conductor (based on its voltage), but the power that's lost as it flows through. That power will be lost due to the conductor's resistance, and will largely be transformed into heat. Heat is particularly pernicious because a heated conductor's resistance will increase, which leads to a positive feedback loop that will result in failure, quite possibly of the spectacular variety. If you have a circuit that draws 30A (typical for an EVSE - an electric car charger), and then introduce a resistance of a tenth of an ohm, the power dissipation over that resistance will be 90 watts - three times the power of a typical soldering iron you might use for PCB assembly!
This PCB trace width calculator is particularly helpful to figure out the balance of current flow, voltage drop, temperature rise and the like. Just keep in mind that at some point, if want to deal with really high current, you're going to find that PCB traces just aren't the right tool for the job.
Anyway, this is just my own observations and experience from the high power designs I've done so far. I don't claim to be an expert, and if anyone sees anything in the above that's in error, please let me know. I welcome all corrections - it's always a learning opportunity.