IF AppleInsider gets the data. You should check the CBC investigation report (available online) which proves that repairable issues are sold as "better to replace your Mac" by staffers. And fighting the right to repair.
Stats are not heuristics. AppleInsider's 2000 units sample is simply too small. How many times I called Apple Support and they shipped me a replacement unit bypassing the store? No amount of figures you provide, 1-10% does not produce the results you claim. It is not just Casey Johnson, but a legion of Apple users with multiple repeats. Now, Casey Johnson uses hers journalistically, so 500,000 typed words in 2-3 years is REASONABLE. Her getting 3 fails, and all others claiming 4- we are now talking with 30% FAIL RATE BECAUSE MATH SAY SO UNIVERSALLY. 25,000,000 units. 10% = one in a thousand odds. Multiple people reporting the same enters the astronomical level. ASSUME 30% or more AND THAT ASSUMPTION MAKES SENSE.
Second assumption that does not make sensE: USE. I walked in Apple store and found DAY OLD demos already not working. Some, like in the video link below, used as a LG 4k monitor, so it is not used for actual demo typing. Video says it all. Ok, if you assume 1/10 then that should have been the only one. i found 3 out of 8 2018 displayed units having the issue.
The 250,000 is a reasonable extrapolation. then you have to use the industrial failure rate and degradation. If 10% first year, Year 4 is usually usually x300% so 10% year one - 30% year 4. However, absent data and the fragility of the system, am not willing to venture such low figures, e.g. 30%.
But no matter what you believe, if users come complain that they get, with reasonable use, 3-4 replacement keyboards in months, the failure rate for that batch and usage type IS 30% or higher. Not what you think, sample or not. It is an industrial degradation and reliability benchmark. Same way NO ONE leaves a car lot with THREE LEMONS IN A ROW FOR THE EXACT SAME ISSUE. If SO, the entire batch was messed. Now, if only 10% of Macbook users actually type as I do or Casey Johnson, that is a different story. So we are now talking 0.3x 2,500.000= 750,000 units per year. But the 10% makes no sense, the assumption is not mathematically cogent to explain repeat failures in identical circumstances. Or me walking in Apple stores and finding new demos with unresponsive keys. What is the usage then? It resting on the table?
IF it was 10%, that is actually industrially acceptable (maybe not Apple and not legally for the price). 1 percent Apple would be laughing. At 30%, Apple institutes a massive recall. And, as I pointed out, even the low usage 12" unit has users reporting failures.
If 30% of users use the pro's solidly for school, then it is 2.25 million defective units/ year. Still a lot.
Anyway, video below. New, unused unit. Apple staffers pretended "We do not get it, it is never used!" lol. I do. Math is fun!