Rdb documentation indicates that RMU LogMiner (RMU/Unload/After_Journal) and RMU AIJ backup can interact in a rare and unpleasant way. The result is the potential loss of multiple transactions to the LogMiner. This problem has been experienced by at least two JCC Customers.
This occurs when the LogMiner starts processing in backup AIJ files and then switches to the live files while backup is progressing. The result is that since the LogMiner is behind and part way through the journal, it is possible that the backup process can finish and erase the AIJ. The LogMiner finds empty space in the journal (the AIJ architecture allows for empty pages) and skips over space that once contained real data and potentially to the end of the journal.
All of this is done silently with no hint of an anomalous situation and so customers can loose transactions in the stream of data going from the database to the Loader target.
The situation is compounded by the fact that the Loader "remembers" where it is in the journal. This memory is contained in the checkpoint record for the Loader family session. Once the checkpoint is advanced, it is advanced. The only solution is to "rewind" the Loader session and force it to start at the beginning of a backed-up AIJ file set. Recall that the earliest of these backed up AIJs must be at a quiet point boundary.
Loader customers should be aware of this and they should be careful when restarting the Loader near the time when AIJ backups are running to ensure that the contents of the journal are not removed.