There are a few areas that could be improved. We have experienced cold start latency because when scaling from zero, the first request is slow. It also impacts user experience for real-time applications. The improvement we require is faster cold starts or smarter pre-warming options. There are also limitations with long-running workloads. It is not ideal for very long or always-running processes, and request timeouts are somewhat restrictive. The improvement we want is better support for persistent workloads without workarounds. We have also identified improvements needed in debugging and observability complexity because logs and monitoring are available, but they feel scattered at times, and debugging distributed microservices is harder. The improvement we require is more unified and developer-friendly debugging tools.
