If a pushState has just occurred and the Safari poller checks to see
if the hash has changed before the hashchange propagates, then it
sees the new state as being different from the actual state, thus
triggering a pushState. This patch sets a flag (waitForPropagation)
which is cleared by an interval after the hashchange finally
propagates.

The question.. what happens if a second pushState occurs before
the hashchange propagates?

History.storeState was only updating the urlToId mapping when new states
were stored during the session, and then updating the idToState and
stateToId mappings on unload, so when the safariStatePoll attempted
to check if the current state was the lastSavedState, that function's
dependency on extractState would fail as it was using the idToState
mapping.