To check my understanding: I believe what is happening here is that we're trying to iterate over all incoming/outgoing edges incident to either vertex 3 or 4. This first means that we have to load the shard that contains these vertices' incoming edges; this is shard #2 in the above example. We then have to find all outgoing edges from these vertices; these will be in contiguous memory in every applicable shard (possibly all of them).
To check my understanding: I believe what is happening here is that we're trying to iterate over all incoming/outgoing edges incident to either vertex 3 or 4. This first means that we have to load the shard that contains these vertices' incoming edges; this is shard #2 in the above example. We then have to find all outgoing edges from these vertices; these will be in contiguous memory in every applicable shard (possibly all of them).