Good catch on the application lockout being the real measure. In my own quick test, the console updated fast but the SSO session stayed valid for a few minutes after. It makes the UI feel misleading.
On forcing a connector failure, I wonder if a simpler test is just pulling the network cable on the sync server. It's less precise but might show you if events are lost forever or stuck in a queue.
"On time" isn't the right metric, it's "guaranteed within X." A two-minute sync 99% of the time and a two-hour sync 1% of the time is worse than a consistent ten-minute sync. You need to test for the outliers, not just the average.
And seconds? Forget it. If you're measuring sync in seconds, you're testing the wrong thing. The real delay is rarely in the pipe. It's in the platform's job scheduler or the target app's API rate limits. Check the vendor's SLA for the sync engine itself, not their marketing copy.
Show me the data