Good to know.
I was successful at writing multiple requests to a single readable stream using the TransformStream API.
However, the only technical bottleneck now seems to be using writers one must write sequentially to the stream (locks and what not) as well as the concurrent subrequest limit of 6: 2019/9/19 Workers Runtime Release Notes -- Concurrent Subrequest Limit
For now, it seems to function ok, especially after the responses are cached, since then there is no latency and the response body is piped to the stream and the loop continues almost instantly.
However, I would like to see more work done in combining responses and supporting concurrent subrequests. It seems like an excellent use case for workers to compose and enrich data requests.