That's essentially the process I used for an application which had a hierarchical graph-like database structure. The naive implementation would check if a node

That's essentially the process I used for an application which had a hierarchical graph-like database structure. The naive implementation would check if a node could be pruned (i.e. if it had no connections) every time a connection was removed. But it turned out it was a lot more performant to just periodically sweep for nodes with no connections like once a day.