I think they've tried to improve error messages and timeouts, and they've done so to some degree, but it could definitely be better. Also my team hasn't use the alerts-as-code system yet, but neighbor teams have, and that seems like a neat addition. The fact that you can generate the YAML in the web UI and just paste it into your codebase is a nice addition.

I think proactive alerting has become more accurate with fewer noisy incidents. lineage and root-cause analysis feels faster and more intuitive now . good balance between powerful enterprise features and usability for engineers * dashboards and observability workflows feel much more polished recently

I like the ML thresholds in Monte Carlo. It automatically trains and adjusts with trends, which is really helpful for my work.