Currently mutation id is an atomic uint64, one per data instance, and is reset during server reboot, so no store is required for every mutation. Given that we are now using mutation id not just for temporary syncs but for kafka log analysis, the mutation id should be considered a property like Max Label. Probably should be an atomic uint64 per repo so that it also serves as an ordering of mutations done in the repo. Double-check use of mutation ids for mutexes before making the change.
Currently mutation id is an atomic uint64, one per data instance, and is reset during server reboot, so no store is required for every mutation. Given that we are now using mutation id not just for temporary syncs but for kafka log analysis, the mutation id should be considered a property like Max Label. Probably should be an atomic uint64 per repo so that it also serves as an ordering of mutations done in the repo. Double-check use of mutation ids for mutexes before making the change.