A while ago ago, we tried launching risky new features in gopls behind experiments, with the prerelease opting in to all experiments. We've since learned that it's much better to simply ship new behavior in the prerelease, encourage usage of the prerelease, and allow for a longer soak time. This leads to fewer moving parts, simpler configuration, and easier testing.

Especially now that we're on the verge of release automation and prerelease canary analysis via opt-in telemetry, we should remove the remnants of experiments.

CC @adonovan @hyangah