The max_snapshot_bytes_per_sec and max_restore_bytes_per_sec repository settings (and various other node-level settings) may limit the speed of snapshots and restores. They work like other throttles: we briefly pause the relevant thread to reduce the data rate to the configured limit. Internally we keep track of the cumulative length of these pauses (separately for snapshots and for restores) but these statistics are not reported in any APIs so it's hard for users to tell if the throttles are doing anything or if there's a bottleneck elsewhere. We should expose these stats in some appropriate API to make the throttling visible to users.