This is related to
Space tracking for Open Containers (committed space in the volume) relies on usedBytes in the Container state. usedBytes is not persisted for every update (chunkWrite). So on a node restart the value is stale.
The proposal is to:
iterate the block DB for each open container during startup and compute the used space.
The block DB process will be accelerated by spawning executors for each container.
This process will be carried out as part of building the container set during startup.