We added additional tags to our web worker metrics to give us visibility into uneven usage caused by (for example) session stickiness at the load balancer level.
We added additional metadata to error reports when individual commands or notes crash to improve our ability to stay on top of those errors.
We improved release notation within our internal systems, allowing support staff and engineers to better see if regressions are caused by a release in the time immediately following that release.
We increased the accuracy of our request/response durations by accounting for time spent doing request logging.
We added the ability to track the request backlog for all web workers, allowing us to track the size of the request queue over time.