Writing audit log entries is very slow.
Even if we submit to the audit backend a batch of log entries, the slow part is to get the entry id based on a sequence for each entry.
This creates a roundtrip with a es/redis or db depending on the UIDSequencer impl for each entry.
This should be refactored to prevent this.
- Extend UIDSequencer to server a block of sequence ->
- For the stream implementation (the default) use an entry id based on the record offset.