The intended behavior was always to block. Dropping data was one of the problems with the original remote storage that we intended to solve with our integration design (go/stackdriver-prometheus-ext). This code was adapted from upstream Prometheus server, and the comment came with the original code.
As the intended behavior is to not drop data, I think we should remove the metric dropped_samples_total. no place in the code increase/decease this metric either. What do you think?
The intended behavior was always to block. Dropping data was one of the problems with the original remote storage that we intended to solve with our integration design (go/stackdriver-prometheus-ext). This code was adapted from upstream Prometheus server, and the comment came with the original code.