In the WM case, could include all steps in an evolution, then draw hyper-hyperedges, which join hyperedges from different steps.
Orange edge represents the updated subpart; which then “pulled back” into the full expression [[[ the rewrite edge could point to the
Orange lines are nodes in the causal graph.... [as are dotted orange lines] Connections: 1. The incidence of orange+ edges [i.e. common expression] 2. An orange line and its “parallel” pullback event [[ every pullback event has a parent, which could be another pullback event... ]] < causal invariance: may have to merge pullback events ... >
[[[ This gives a bug ::: ]]]
Second type of edge joins tree-like separated events..... Probably the true graph removes some of these events....