Here at our Fort Collins, CO site we've implemented 3 Linux HA clusters so far.
Cluster 1 is a middle ware Oracle Database and Web server that was implemented when the native Oracle clustering tools failed to provide smooth failover. At the time of deployment, Oracle RAC-9 was a new product on Linux, and turned out to have some early adoptor issues. We switched to a dual active database and Linux-HA to keep the project moving. The cluster has been running for nearly 2 years. Cluster 2 was built as a second middleware set of Web servers.
Both Cluster 1 and Cluster 2 track and report realtime status data for our manufacturing process line. Availabilty of the realtime status data depends on the virtual server to be available to collect, process and present (via the web server) the data to management as well as operations.
Our third cluster is used as our site's e-mail master host. Using the virtual host to float between the two nodes allows for seamless avilabilty to the users while we perform maintainence. Due to the volume of e-mail, we would frequently swamp our upstream e-mail hop with queued mail if our old server was down for several hours.
Heartbeat is being tested for deployment of cluster nodes separated by building at our site. To avoid single room power failures (due to scheduled bi-annual system inspections) we are using Intel fiber LAN cards to drive heartbeat over 50micron fiber links between our buildings.