At work I also used to stare at a lot of graphs to figure out bugs. We reuse some of the graphs so we've gone through the effort of recording node positions in metadata and given the graphs a UI where people can drag nodes around and re-save. But it's clunky.
I always wished I could invent some domain-specific heuristics to lay out the graph automatically but never took the time to.
Manually moving nodes actually worked fairly well for our use case, but it made a mess of text diffs when e.g. nodes were inserted and the position of all subsequent nodes shifted, so as a consequence we had to invent ways to visually diff graphs. Maybe the improved automatic layout would have been easier after all.
I always wished I could invent some domain-specific heuristics to lay out the graph automatically but never took the time to.
Manually moving nodes actually worked fairly well for our use case, but it made a mess of text diffs when e.g. nodes were inserted and the position of all subsequent nodes shifted, so as a consequence we had to invent ways to visually diff graphs. Maybe the improved automatic layout would have been easier after all.