During graph initialization, the master and slave graphs start at the same time.
If the master graph activities block during initialization, e.g. because it is trying to write to a file that is locked, then the slaves will eventually timeout with a "Error receiving actinit data for graph: " message.
Not new behaviour, slightly surprising not seen before.
It looks like behaviour changed between 5.6.8 and 6.0.x.
Some refactoring changed the execution order subtly, previously it would have initializaed the master graph/activities, before sending the slave to the graphs and consequently starting them and the timeout countdown.