To achieve atomicity we lock the container path.
However, the write operations update the statistics for the ancestor
nodes (container/account).
Therefore, the backend restricts the recursion up to the container
level.
Account statistics should be computed whenever are required.