But I will never forget, early in my career visiting a potential vendor to make parts for a product I was working on. One of their jobs was chemical milling on major structural components for the Space Shuttle. They commented that where the loads on the structure were well understood they were using a 10% safety factor, but when they were less well understood that went up to 20%.
Yeah, aerospace can get insanely tight on margins. When full operating history of the member's loading is known -- whether by direct measurement or modeling -- you can pull margins very close indeed.
And the Space Shuttles were never intended to fly, all that many times -- I don't know what they used, obviously they did make repeat trips, but I doubt they expected more than a thousand say, which means fatigue limits can be quite loose. Not so much for rapidly rotating or vibrating equipment, like the rocket engines -- turbine parts can be tuned for single-digit failure points IIRC(!!) -- and those may indeed be "wear" parts, to be replaced every flight, or regularly scheduled in any case. Ditto anything like, drag race engines, say.
Likewise, the materials and processing are very well known -- or can be, when there's economic incentive to do so. Just as we can spend inordinate amounts of time researching every minute detail about say ceramic capacitors, so too can the whole industrial process, from melt formulation to microstructure and working/forming to finish machining, be examined extremely closely.
Conversely, when it's not, it's not. A36 architectural steel is somewhat notorious as an alloy, because it's....not, it's basically one thing: 36ksi minimum yield. It's cheap, which is also to say: highly recyclable. There are some limits on elongation, ultimate strength (min/max), and chemical composition (max of several common elements), but a real sample could contain a fair mix of minor alloying elements, and make quite a metallurgical surprise if you [mis]used it for something technical like tools, or, engine parts or something. Ditto rebar (which usually has a little carbon, giving higher strength).
Mind, not that A36 can nominally be hardened -- below about 0.4% carbon (it's under 0.3%), you have a very hard time quench-hardening steel at all, let alone to the degree needed for tooling (which needs more like 0.8-1% for plain carbon steel). But, one might start with random scrap steel and then case-harden it, or a blacksmith might use scrap as a backing, forge-welded to a strip of tool steel for the tip/edge; etc. Surprise alloys might not matter in normal handling (i.e., bolted and welded structural steel), but once you start doing metallurgical treatments like these, it starts to matter quite a lot. And when that tool shatters in the quench bucket, after a couple dozen hours pounding away at it... you might just want to blame the alloy, and, rightly so to a fair extent. (A better craftsman of course either understands and accepts the peril -- like a responsible EE gritting their teeth and selecting Z5Us when cost/density matters and the value spread can be guard-banded, or eventual failure is just less critical e.g. outside of warranty -- or one selects better materials in the first place, when it matters; why risk making a chisel or hammer out of something that could spall off sharp high-velocity chunks near ones' eyeballs?)
In any case, what you get for common hardware store "mild steel", is probably close to what it says on the label. Generally you can't go down in strength, without it being just literal sponge (slag inclusions, gas pockets, rusted to hell?), in which case, how did they even manage to forge a rod/bar of it at all?! -- but you can go up in strength and end up with lack of flex, or brittle failure, especially after processing like in the above scenarios. And in any case, we're talking like a spread of a factor of 2, say -- like from HRS yield ca. 30ksi to CRS ultimate at 60 -- plus another factor of 2 to 4, on top of accounting for peak loading, yeah you can have something that's essentially never going to fail due to loading.
The direct electrical analogy is designing to the test: if we need to pass ESD and surge, well, put in a device rated for about ten strikes each, and if it survives, great, that's it, you're done! Will the equipment survive longer in service? Well, that's another matter, but it's also rare that protective devices fail like clockwork, and any additional margin on ratings that you've design into the product will keep it that little bit safer. At least until something else fails, like if you put in the SMA-package TVS (ESD "safety factor" of say 4 or more?) but forgot to get a peak-rated chip resistor beside it that happens to blow from a single 10kV strike...
Tim