When you use the audit sink policy, it can greatly increase the latency.
The audit sink is blocking and we have had big latency problems with it.We were unaware of this because we calculated our latency before going to the logsink, but got complaints from clients connecting to us, bringing the issue to light.
The logsink policy was increasing latency by 90ms versus using the internal audit database. We see latencys in a policy with only a return-template response of 8ms without audit-sink-policy, and up to 100ms when sending logging to a backend sink policy. The more detail audit loglines inside a policy, the worse this was getting.
We since switched to a different approach using the accumulate data in memory assertion in the logsink policy, which does work fast. We read from the buffer in a scheduled task at another location.