I'm not an engineer, so you may well know much more about this, but it seems reasonable to me that real-world people, or some percentage of real-world people, aren't going to hit your keys in the exact way you tested for. If the effects don't appear for years out, what type of user test can you design to account for that?
As a product category, laptops have existed for close to 40 years, and Apple has been a successful and high end participant in that category for close to thirty. If they can't produce a reasonable keyboard reliability test or lack the willingness to do so, then they have other problems that are vastly more concerning than anything about a couple specific computer models.
More importantly, Apple did not need to also ship the 2017 MPB, the 2018 MBP(s), or the 2018 MBA. If they didn't have any alternate designs in the pipeline (why is that?), they should have reverted back to the 2015 design, with new parts.
Well, that's the point where I'd call Apple's judgement into question. They should have put the breaks on new releases with the keyboard until they had a clearer picture of what was going on.
It's really not - predicting lifetime expectancy and failure rates is a discipline of industrial engineering. That is why it's possible to estimate the number of open/close cycles of hinges is possible to estimate with a great degree of accuracy. A company has to master art this for selfish reasons - so that they can minimize repair costs during the warranty period