Ed, Tuc,
I may have this wrong, but my take on your posts above would be:
It is usual to consider multiple failures in similar areas - hence the point of zonal anaylsis to look at issues that a straightforward functional analysis of failure modes might miss (i.e. a fuel system functional analysis might not consider consequential failures in a flight control system that is in its proximity etc.). This is especially relevant for the cascade failure modes that Tuc points out where you could have a situation where a fuel leak (e.g. prob 10-3) might saturate a theoretical flight control system component and make an aircraft crash. This would be a cascade failure where one not unlikely event (10-3) would cause a definate catastrophic event & hence wouldn't be allowed (though of course no flight control system would ever be designed that was susceptable to such a single point failure). This situation is similar I think to the "Nimrod hot pipe issue" in that fuel spilling on a 400C pipe will ALWAYS cause a fire and is therefore not acceptable.
Edset is right however that you wouldn't normally rule out combinations of events where the cumulative probabilities are infact low, but Tuc is right in that you should consider them. In the electrical spark scenario, you can say, e.g that prob of a electrical spark fault is say 10-4. Now this can't be allowed in a fuel tank where volatile vapours are very likely, but in a bay where the prob of a fuel leak is 10-3 this potentially would be allowed as the cumulative probability would be 10-7. It is worth noting that 10-7 would be well into the broadly acceptable category so the risk wouldn't even need to be reduced to ALARP - it already exceeds the acceptable risk threshold.
I've made the probabilities up - they are only to show the principle of cumulative probability. The risks of electrical 'sparks' in a fuel tank (TWA800 etc.) should not be confused with those in a normally benign environment.