That's a somewhat misleading term! I much prefer "probability one".
> But I guess here you mean something else and I'm not sure what.
Read it literally. The everyday definition of "almost". In other words, "the vast majority of the time".
“probability one” is misleading, too, as laymen would assume the reverse cannot happen.
The interpretation of “almost always” as ”p = 1 despite the existence of possible outcomes that defy it” likely is based on http://mathworld.wolfram.com/AlmostEverywhere.html (which has “almost surely” as the mathematical expression, but I would immediately read “almost always” with the same meaning)
In the real world it cannot happen. It can only happen in an unbelievably unlikely mathematical abstraction. So I would call that layman understanding closer to the truth.
"almost always" is more like a 95% threshold, sometimes higher but with an implication that failure is realistic.
"Almost always" is a term of art in probability theory, where it means p = 1 despite the existence of possible outcomes that defy it. For example if you keep tossing a (fair) coin it will /almost always/ eventually come up "Heads". But I guess here you mean something else and I'm not sure what.