Actually the doc appears to be saying what I did, so the documentation is correct about this behaviour. Picking IS happening on all conditions, but the effect is cumulative. If you start with 10 instances and the first condition picks 5 then the next condition can only pick from those 5, if that condition picks 0 instances then any following conditions have no instances to pick from and hence are all false.
In the example you gave each condition is picking from 2 sets of instances, the observer and the targets. Because the observer is used in each condition if any are false the following conditions has 0 instances of the observer to pick from, but because the targets are unique to each condition they have the full set. So it's nothing to do with "else", just how "OR" behaves.
I admit this behaviour feels a little weird to me, I would have assumed that picking wouldn't work like this in OR blocks, but I'm not familiar with the history of this. There might even be a good reason it works like this that I'm not aware of, but the design of the event system predates me joining Scirra by quite a few years.
About the reply the website sometimes has caching issues with posts, so they don't show up for awhile or until the thread is updated again. Typically editing the post is enough to make it appear.