I'm currently building a serverless architecture and was hoping to use elastic cloud as a one-stop repository for logs and metrics. Per #320 it seems that Lambda is not a supported environment for Elastic-APM.

I'd be happy to put in some work to make this happen, but I'm not entirely clear on where the problems lie. Is there a reason why this is more complex than calling collect on the metric repository and flush on the transport?

If I had a MetricRegistry with a collect_interval of 0ms, and an unthreaded transport that flushed only on demand, I could flush explicitly at the end of the lambda invocation. It would be simple to extend that to a decorator that flushed on exit from the wrapped function.
Is this an enormous simplification of the requirements?