TL;DR: Dijkstra's alone isn't sufficient to solve the all-pairs shortest paths problem, but on an undirected graph, you're right that there is information contained about some of the pairwise shortest paths (i.e., those involving the source vertex as the source or the destination). Depending on the queries (if you know them ahead of time and they were sufficiently contained in a small subgraph), perhaps, you could run Dijkstra fewer than
Hmmmm... I'm don't think Dijkstra alone would give you all-pairs shortest paths (APSPs) among all vertices inside the envelope[1] as it proceeds (or even at the end). The distance estimates are from a single, source vertex
Johnson's is an APSP algori