Envoy has default settings that can surprise new users and hurt Envoy's perceived performance in benchmarks. For example, I think at the Pinterest EnvoyCon talk (@derekargueta) mentioned the default circuit breaker settings limiting the number of concurrent requests.

We've had reports from folks who have compared Envoy's performance with other proxies using default settings and found it underwhelming. We know that horse race numbers can be significantly improved by disabling dynamic stats, etc.

Should we be making changes to default settings or (more likely) better capture best practices for Envoy production and benchmarking configuration?