Multiple site mode breaks concurrent cluster node bootstrap

It appears that attempting to start multiple cluster nodes simultaneously, with all of them using repo.bootstrap=true and using multiple site mode, will almost certainly fail due to concurrent, inconsistent writes to the HCM baseline.

The obvious workaround options are to 1) don't use repo.startup=true on multiple nodes simultaneously or 2) don't start up multiple nodes at exactly the same time.

However, we do intend to support multi-node-startup, so we need to build some kind of coordination mechanism to handle this.

please see attachment file for the detailed logs.