To better test performance and memory improvements, we need a consistent benchmark to compare changes against.

One idea is to have an AWS environment (spun up with Cloudformation or Terraform) that we can recreate and test against. We can then measure things like memory usage and performance against this environment. This has the downside that it relies on AWS and latency from the test machine to some extent, but the advantage that reflects a very realistic real-world scenario. We can also mitigate some of the downsides by running the test itself on an AWS EC2 instance of a particular size and in a particular region.