I was thinking of those as thing you do before product release (so they're "known"). But it's not a good way to find out about reliability issues, because those only happen in especially weird situations, or over time like running out of disk space.

Telemetry that tells you which features are popular is useful but does need filtering to avoid identifying individual users. But sending back errors and crashes is what's really important.

You can do things like have feedback forms but typically users don't like sending that in because they feel like they're doing work for free.