Description
AbstractNotificationRegistryTest::registerShouldWorkWithExpiracyDate have a slow percentage of chances to fail...
A delay of one second is used. This leads to push a 1 second TTL in Cassandra.
A quick experimentation seems to suggest TTL then expires when the next second is reach (Polling for checking expiration delay suggests we got values between 0 & 1s)... Which of course is hasardous, and lead to random test failures.