It has been decided a useful variant of the replay functionality would be to replay all requests as quickly as possible.

The intent of this is to apply a real workload, but artificially "scaled". This could be used to give customers confidence on e.g., the peak workload the current deployment could handle.

Initially this need not expected to perform any complex data dependency analysis to keep request ordering consistent - instead sessions will be allowed to de-synchronise, and the inaccuracy of this replay is accepted. This may be re-assessed in the future, but for now the simpler approach allows testing and investigation.