Since I am not a trained engineer or fabricator, I built using components that I calculated had a high safety margin for the loads they would see. But, there's hand cut threads here that need to support pretty high loads, and a lot of welds that might not be as good as they look (though a couple professional welders have given it thumbs up).
My solution was to actually load the assembled structure to twice the max load that it could be expected to encounter. According to most motorcycle builders, this means 1g of braking force, and a 3g 'bump' force coming through the suspension to the frame. I figured I would do this testing before bothering with any steering mechanism, because if it failed the tests, steering wouldn't matter. However, I needed to have my load bearing structures built to pretty much the form they would be in for actual use, which means I worked on the bike for nearly two years (though only a few days a month) before doing these potentially destructive tests!
To simulate the braking force, I built a long lever who's fulcrum was the axle. It applied its force to the fork in the same fashion brake caliper mounts would (via machined blocks) and was long enough that the load on the end was multiplied by 4 as a simulated brake force. I then suspended 375 lbs from the end of the lever arm, creating a simulated braking force of 1,500 lbs - double a 1g braking force for a bike with a gross weight of 750 lbs. I also used some dial indicators and mag mounts to measure the amount the fork was bent back, and the amount the frame itself was bent back, allowing me to get an idea how stiff the structure I added is. Turns out is it very stiff indeed, total deflection during this test was under 0.25 inch.
Constructing a lever setup to do the bump load tests was a serious design challenge, so I bought a 1 ton crane scale and used a turnbuckle to apply the load. Each axle got 1950 lbs pulling up on it, with the shock replaced by a ridged strut and the motorcycle frame held in the jig via its engine mounts (as for the above test). Both resulted in a fair bit of frame flex (no surprising given the diamond shaped hole in the middle of the frame), but the load remained stable, indicating the loads never got past yield strength. On a racing bicycle this would be considered desirable "vertical compliance".