And it's a good bet that improper handling with no static-discharge precautions at any point in their lives is a contributor to their failures, too. Based on my experiences at Honeywell Commercial Flight Systems Group in the late 80s/early 90s, and all the problems we traced down to ESD, most of them caused by people who knew better but still didn't follow precautionary guidelines (both within our plant and from external vendors), and some of them caused by vendors that didn't have any kind of precautions at all, and no training for their workers (regular tile floor with plasticy coatings that you could *feel* the static buildup from, if wearing normal tennis shoes, wooden or plastic-topped benches, workers in street clothes or worse in nylon lab coats!, the list could go on.
ESD doesn't usually kill something outright, it just sets it up for failure later under stressful conditions. If being inside a hot motor with all sorts of changing conditions around it, electrically and otherwise, isn't stressful, I'm not sure what would be.
In the throttles it's more likely to be a mechanical stress, but can be water, sweat, dirt building up on it and shorting the leads out (or capacitively allowing current flow between them), voltage spikes or fluctuations from poorly regulated outputs from the controller or induction along the long wires from it, etc.
If the parts are "seconds" to begin with, or even outright fakes (other lower quality or rejected parts rebranded or restamped with a "better" spec'd part number), both of which are common enough in today's market for semiconductors, it would be no surprise to me if either or both of the above problems caused a lot of the failures seen on bikes, in all types of electronic components.
It's unlikely that most companies producing things care much where their parts come from, or do much to validate their origins and authenticity, given the huge problem the capacitor plague has been for NEARLY TEN YEARS. If the distributors that carried the parts were to test and validate their parts, and refuse to buy from vendors that repeatedly provide defective or non-passing parts, those vendors would either cut down or cease their problematic methods, or go out of business. But because parts are cheaper from places that dont' do as much (or any) validation, that's who they buy from, forcing the places that did do it before to reduce or stop their efforts to compete pricewise.
If the manufacturers of FRUs and TLAs (end-products) would refuse to buy from distributors that provided bad parts because of lack of validation, then the distributors would be forced to do the testing or themselves go under. The same thing applies here as above, that instead they'll buy the cheaper parts and eventually cause the opposite of what needs to happen, by making it so NO ONE validates anything, reducing the whole chain's quality control to almost nothing.
Ideally, consumers should do the same thing--buy only from sources that validate their products, but I already know that consumers aren't going to stop buying from manufacturers that repeatedly provide bad products, because consumers are even cheaper than companies are, and more willing to live with crap for a while until it breaks and they can now go buy a new one from another crappy cheap company.
It's been getting more and more like that over the last couple of decades, and the economy (at least in the USA) is getting to where consumers can't *afford* to demand the best, by paying the premium prices for everything. Many can, still, but more and more people have to cut back more and more, and it's going to get a lot worse before it gets better.
In the end, the only parts that are really rejected outright are the ones that are simply dead off the original production line at the component manufacturer. Even then, I expect a significant number of the lower-end places will put as many of those dead parts as they can into batches of working parts, just so that they won't have wasted as much money. They know that not much testing of the parts will happen after they leave their hands, right up until it's in the end-user's hands.
I guarantee you that with many, perhaps even most products, a majority of any particular batch of them is not even powered up before it leaves the factory--I have repaired so many things over the years that could NEVER have worked, with caps, diodes, transistors, and other highly polarity sensitive parts in backwards, and even with wires missing from the power inlet to the PCB. Impossible that it was ever checked--just thrown together, stuck in a box and sent to a store to be sold. Not all this crap comes from China or Asia either--some of it is US-made, some of it from Canada or Mexico. Not much of the later three, most of it Chinese or India or other such places, but enough of it to tell me that many companies just don't care, or else they simply can't afford to care anymore if they are to compete pricewise with the places that really don't care.
So I think that the least reliable ebike component, aside from the operator (Zoot Katz is right), is everything else.