When you have different manufacturing facilities you will always have this issues even when you are building to the same spec. Your manufacturing process will have different tolerance biases that's unavoidable. The problem happens when you stack tolerances in an additive manner for example when you have a nut which is built to spec in plant one and is 0.005% too big (which is well within the tolerance limit) and you have the bolts built in plant 2 and they come out 0.005% too small (which on their own is still within the spec) you get an overall tolerance shift of 0.01% which all of a sudden is out of spec. This just happens in manufacturing all the time the parts are in spec individually but the end result just doesn't work.

I actually deal with this all the time in hobby electronics you buy various parts which all of them have a spec with +- certain percent and you when you buy them from a single batch you can be quite out of luck and find out that when you add them all together the tolerances just stack against your initial design.

Speaking as an engineer, you're supposed to design for the worst-case tolerance stack to be viable...

If 0.01% is out of spec then you need tighter tolerance specs on your individual components or otherwise your design does not mathematically make sense.

This sounds like a design and process control issue as much as anything.

It is a process control issue, the problem is that controlling the process when you have 100 of sub contractors manufacturing parts which are pre-assembled in 10 different locations and then sent for final assembly isn't that easy. And even if you add tighter tolerances to the spec you can still end up with things not working properly.

Not to mention that increased tolerances means higher costs and there's a point where you can't adhere to them physically and as far as real world manufacturing goes the tighter your tolerances are the tighter the control and QA needs to be which results only in diminishing returns.

When you end up manufacturing things tolerances can be skewed by a 100 different reasons from a slightly different zero and certification process to different CAD/CNC software which rounds up thing slightly differently. The ISS development actually learned quite a few things with that and they've both adjusted their process segmented the manufacturing by entire modules rather than individual parts then the only thing you need to really worry about are the tolerances for the actual dock which have quite a bit of wiggle room. If you had to make the frame for the cupola in Italy, the outer housing in France the Windows in Germany, the Shutters in the Netherlands and assemble them in the UK you would ended up with similar problems.

Saying there's nothing wrong with X you just need a better process can be said for pretty much everything but that argument rarely holds water when you deal with real world applications, there's a reason why there are entire engineering disciplines for manufacturing and process control.

So no as an Engineer you need to design tolerances not for the worse case scenario but for the actual manufacturing process you are using and the controls you can enact on that process. When that process is then spread out across as many parties as possible it's no longer the process you've started with and quite likely isn't neither and ideal process nor one you could ever really optimize.

There's a good reason why some of the worlds most advanced engineering parts are still quite often hand fitted at the end, and that's when you work with a single manufacturing process in a single factory. When you spread that out you end up with parts that sometimes can't even be hand fitted anymore.