Why does the lambda container has to pull events from the lambda runtime using an HTTP GET request, instead of having the runtime push the events with an HTTP POST?
This is all based on how the RunTime API worked already (pre-dates this launch by 2 years or so). We wanted to not change too many bits.
Since function code has no listening socket/port as it were you need something (like the bootstrap script in a custom RunTime) to pull the local interface for the runtime API. That runs on the underlying worker and communicates with the API for Lambda.
How would the runtime know for sure when the container was ready to accept the request? The lambda model--with the container pulling the events instead of them being pushed to it--seems like the correct way to model the concurrency.
