Cluster Safety in a Single Node Standalone Instance
Why have Cluster Safety in a Single Node Standalone Instance
The reason the cluster safety mechanism exists even in a standalone installation is because it's possible to have two Confluence servers on different computers, both pointing to the same database. For example, in creating a test server, if you copy the data over from production but forget to change the database information, you would have two servers accessing the same database, and it's possible to corrupt your data.
In those circumstances, the cluster safety mechanism will kick in within a maximum of 30 seconds, so the likelihood of corrupting data is almost none - Confluence will still be starting up in the first 30 seconds and is mainly reading from the database, so it's very unlikely a write could occur before the instance shuts down. If you were to increase the cluster timeout to 300 seconds for example, it would be possible to run two instances by accident for up to five minutes, which is enough time to do a reasonable amount of damage to your data.
The way cluster safety mechanism handles a cluster panic differs depending on the Confluence version. For Confluence 5.4 and earlier, the cluster safety mechanism will shut down all instances. From Confluence 5.6 onwards only one instance will be shut down in a cluster panic.
When a cluster panic occurs in a single node environment, it's usually possible to rectify it by restarting Confluence. However, if that is not successful, there are some other potential causes and resolutions outlined in this document: Cluster Panic due to Multiple Deployments