kore-repl keeps an explicit representation of the entire execution graph in memory, but it is not necessary for kore-exec to do so. Keeping the graph resident causes excessive memory use on large proofs. It is only necessary to keep the graph's leaves in memory.