Earlier I said "Products that don't incorporate [telemetry] are going to be clobbered by products that do". High QA costs are the biggest part of this.
As it should be when you're developing an OS for 90% of the market across a huge variety of hardware. There's really no way around that, and jettisoning QA teams as a cost-saving response is incomprehensible to me.
Trusting telemetry alone to ensure the robustness of your product will result in a "good enough to ship" focus on mainstream work-flows and leave corner-cases unaddressed since they're rare. And corner-cases are where the bad, system-breaking things happen.
Telemetry should help QA people write mundane automated test-cases to cover the 99% and then enable them to go and spend the constructive part of their time focusing on breaking the system. But it's absolutely not a replacement for QA.
From where I sit, both roles are combined. The output is a far smoother engineering process and a higher quality OS. You mentioned 'corner cases [being] unaddressed' but I've witnessed the exact opposite: most of the criticism I hear about Win10 has nothing to do with quality.