The idea is to work out the maximum design limit, then add a factor of 2x or 4x to that to allow for things you can’t predict or model well. If the design is relying on some part of the above of the 2x to work, it’s not really a safety factor.

I’m a mechanical engineer and figuring out the appropriate strength a part needs, then making it twice as strong, is a common tactic. You can’t account for material differences, little casting voids, or just your math being off a little by the approximate methods we use to calculate stresses. The more perfectly we can model and then repeatedly destructively test a part, the closer we can get to a safety factor of 1.