Actually, all standard failover cluster products I know just 'feel'
fragile. If You want high reliability, You need to run Your
applications really parallel, not failing over to another host if the
current host fails, but run it on all hosts at once, like it's done
i.e. in high end telco systems. I doubt it's easy to implement, and
You'll probably be better of by toughening HACMP.

I dare say that because after it I'll point You to
http://www.infrastructures.org - give it a look, they say they're
setting up full HACMP clusters in two hours without any manual
intervention. I think that's robust enough.