There was no Telemetry outage. An HTTP3 response header’s case was changed by a third party without notice. Telemetry continued working, other than the case change causing a bug.
There was an http3 infinite loop bug in Firefox that crashed all networking. Many different things could have triggered the bug once it was introduced. Telemetry happened to be the first thing to do so, but not due to any faults in Telemetry’s code or implementation.
There is a why here, and it includes Telemetry mixed traffic as a potential culprit. There are reasons to unify traffic (proxy support, QoS and whatnot) but unification of the user and Telemetry streams isn't without risk, as has been shown.
A constant refrain over the last 10 years or so of Mozilla's descent while trying to justify the removal of features from Firefox has been that not doing so unnecessarily bloats the surface area of the codebase, and specifically that this increases the chance of vulnerabilities and defects.
Will the same argument be applied here, now with a case in hand, to justify the removal of telemetry, too?
What do you mean that's not what happened? Was there a "Firefox outage" or not? Are you disputing claims that the engineering team made about Firefox becoming unusable for users "for close to two hours"?
That's my point. A "telemetry bug" didn't make Firefox unusable, a networking bug that was triggered by a telemetry bug did. But it could just as easily have been triggered by anything else.
Secondly, "cause" doesn't automatically mean "root cause". (That's the entire reason we distinguish between the two by qualifying the latter to begin with.) It's perfectly reasonable to say "A caused B" even if the root cause lies elsewhere, with C.
Thirdly, none of this matters. It has no impact on the point being made by the person you responded to, which—to repeat—is that:
> It should be impossible for the phrase "the recent Firefox outage" to make sense.
It makes perfect sense a world where half the internet is going through Google / Cloudflare / Amazon /Akamai servers or some combination of the above, and they decide to roll out brand-spanking-new protocols to half of the internet at once. Sometimes that's going to break clients.
I don't like that world very much, but it's the one we live in.