PubNub's auto-reconnect and catchup mechanism is designed to handle brief network interruptions (e.g. driving through a tunnel, switching from WiFi to cellular, from one cell tower to another, etc.). That mechanism is effective whereby a resuming subscribe is able to deliver published messages that were missed during a dropped network connection.
The subscribe in-memory queue is volatile and serves as a high-performance message cache utilizing optimized system resources and subject to garbage collection cycles (5+ minute garbage collection cycle). Hence, it cannot be used for long-term catchup purposes.
The default message queue size is 100 messages. Consequently, publishing past 100 messages inevitably results in older messages overflowing the queue and getting discarded. This should be kept in mind as another constraint of the in-memory queue in situations involving a fast publish scenario. If you had published more than 100 messages since the subscriber went offline that offline subscriber would miss messages that got pushed off that queue upon resuming the connection.
To guarantee message availability via access to historical messages, we always suggest using the history API which works in tandem with an enabled Storage add-on feature.