There are robust ways to do this. The fact that some choose to implement their own buggy protocols for it does not mean that I do not want it. I use automatic failover with SQL Server today (and has been for years). I don't like the idea of getting up in the middle of the night to acknowledge some hardware error to initiate a failover.
The fact that a commercial software empire with tens of billions of dollars in revenue can, in conjunction with an OS over which they have complete control, implement working resource fencing does not imply that a mostly-volunteer Open Source project, running on at least a dozen-odd different hardware and software platforms, can too — and certainly not on the same time scale.
