Make testRetriesEventuallySucceed more reliable (#1779)
Motivation:
`testRetriesEventuallySucceed` fails infrequently. When attempting a
retry, the executor checks whether it's safe for the broadcast sequence
to be replayed, it does so after removing all subscribers. However,
there is a timing window between invalidating subscribers and checking
whether it's safe in which a subscribtion can be registered. For
retries, this can happen if the server responds before the client has
subscribed. In practice this is very unlikely to happen because of
network latency.
Modifications:
Have the server in the test consume the inbound stream before failing,
this ensures the subscribtion is created by the time the client realises
the call has failed.
Result:
Test passes more reliably
George Barnett