JLSeagull,
Semantics can be important. "Fail Safe" as an engineering concept is not important, and is often worse than worthless. Designing a system that fails in the safest possible manner (taking into account both personnel safety and protection of equipment) is very important.
The analysis you alluded to above is exactly what needs to be done. The designer needs to look at every credible failure and determine how he wants actuated valves to perform during that failure. Every ESD that I've ever installed has been pressure to open, pressure to hold open, loss of pressure to close (I've always done it with the huge-spring kind of actuator). That way you don't have an operator standing in a fire trying to pump a manual actuator. But that is just one facet of the analysis. How do I want the solenoid that supplies gas to the actuator to fail on loss of signal? Loss of power? Does it need to just close or does it need to vent downstream? I say that the answer to every one of these questions is valve-specific and the PHA (or one of the less formal alternatives) needs to address all of these questions instead of being a PSV-Minette that is so scripted that no one would consider asking a question that was not on the script.
David