Daniel Passos Paolo Bueno David Martin There are 2 different ways to implement this:

1. In the executors we keep an array for the results, and when all the checks are completed, we call metrics.publish once to send all the metrics
2. The metrics service itself will handle the batching. That means the metrics service itself needs to keep a buffer somewhere, and send the metrics data based on certain conditions

No 1 is easy to do, probably should be the solution for now. But I think if the metrics service is something that we want to keep long term, and it is really becoming a metrics service for arbitrary metrics data, then we should look at No 2. It will also help with failed requests etc.