For decades CR has been consistent and, in statistical terms, appeared to be both valid and reliable. That said, to more complex the product tested and the more unconventional the product tested, the less the actual utility of their processes.
For decades we all knew their reviews of Porsche, even Mustang and Camaro, would be less useful to potential buyers than would be if the comparison were between competing minivans, for example. So, where does Tesla fit in?
Tesla is anamolous in so many ways that it broke all the CR paradigms. First it was the "best car ever tested" then the Model X was the "least reliable". These extremes are entirely understandable once we consider who CR are, and how they approach testing and surveys.
CR testing is mostly about statistical comparisons of performance, as in " how much damage does the high spin cycle cause for delicate clothing". By the numbers the Tesla Model S blew away the competition, except for interior space. As for missing handles to help Ingress/egress, adequate interior storage and all those crucial minivan attributes, even CR's normallly phlegmatic testers were overwhelmed.
CR reliability data makes no material distinction between problems identified and fixed remotely by a manufacturer and any other problem. They make little disntion between a dashboard squeak and a transmission failure. A driveline replacement made prior to failure rates the same as one that left the driver stranded. If there was a free loaner car during repairs, doesn't count. Then, intention to buy another, and owner satisfaction does count, but only completely separate from the JDPower/CR style, Things Gone Wrong Metric.
Nobody needs to game the CR system to produce these results. They only need to ignore the metrics actual owners value when assessing their own satisfaction. CR, in my opinion, is quite honest, but they are limited. So what?
For my part I am a loyal CR devotee for refrigerators, washing machines, and other appliances. So, also, for mobile telephone service providers, soap powders and so on. Otherwise, even for smartphones, much less cars, I would not dream of making a decision based on either positive of negative feedback from them.
I sympathize with the dilemma presented to CR, and others, by Tesla. Traditional paradigms are breaking...