Telemetry in Go 1.23 and beyond

From the article:

As Russ Cox pointed out in his original blog posts, a drawback of the off-by-default approach for telemetry is the continuous need to encourage participation. It takes outreach to maintain a sample of users that’s big enough for meaningful quantitative data analysis, and representative of the user population. While blog posts and release notes can boost participation (and we’d appreciate it if you would enable telemetry after reading this!), they lead to a skewed sample. For example, we received almost no data for GOOS=windows from early adopters of telemetry in gopls.

To help reach more users, we introduced a prompt in the VS Code Go plugin asking users if they want to enable telemetry [...]

...

The most surprising aspect of the bugs coming from telemetry was how many of them were real. Sure, some of them were invisible to users, but a good number of them were actual misbehaviors of gopls–things like missing cross references, or subtly inaccurate completion under certain rare conditions. They were exactly the sort of things that a user might be mildly annoyed by but probably wouldn’t bother to report as an issue. Perhaps the user would assume that the behavior was intended. If they did report an issue, they might not be sure how to reproduce the bug, or we’d need a long back-and-forth on the issue tracker to capture a stack trace. Without telemetry, there is no reasonable way that most of these bugs would have been discovered, much less fixed.