We need to be careful not to conflate two different things here. As @iserith has pointed on several occasions, the intention of the passive skill rules is not originally to model passive vs active usage; it's to obviate the need for a dice roll in circumstances where an outcome is "fictionally uncertain" but rolling dice would be inconvenient. The two most common reasons for that are 1) not wanting to reveal the check to the players - such as for monster stealth - which is also a lot less "swingy" when rolled against a PC PP and 2) handling repeated actions without rolling the same check over and over again. Both of these are obviously common in the case of Perception.
But I think there is also a distinction to be made about levels of attention. You can explore the dungeon at a run, at a normal pace, slowly and carefully, or even inch by painstaking inch. Clearly you are *not* going to notice the same things at a run as you would if you spend an hour on each 30' section of hallway. You can simplify this into as a distinction between active searching and passive perceiving; but I think it's a big mistake to map this directly onto active vs passive checks, because they're not the same thing and the maths of passive checks are a very bad fit for modelling active vs passive engagement with the environment.
How this works in practice for me: when moving between rooms and along corridors, I tend to use passive perception to avoid the need for dull, repetitive rolling. The players can choose whether they want to move at normal pace, in which case I don't track time for these transitional moves, and Perception DCs for hidden things are +5; or slowly and carefully, in which case each "transition" takes 10 minutes, but Perception DCs are normal.
When entering a room, this framework helps also me be consistent about what I reveal immediately to the PCs initial scan - which I take to be equivalent to moving at normal pace through corridors. The only difference is that since rooms are the main "encounter spaces", I won't normally use PP if the PCs do a 10 minute search - I'll have them roll because they've declared an explicit action ("we search the room") rather than a background task ("we search as we're going along").
This can be flipped either way where appropriate. If the PCs feel a particular section of corridor is really important than they can make a Perception roll to search it; conversely, if they say "we're going to go room to room methodically, searching every one", I might use PP for everything.
It's important to note that if you're going to run PP this way, you need to be strict about limiting the PCs to either PP or an active roll unless they are committing to a different level of search at the cost of more time. So they might walk down a corridor, noticing things with PP (DC + 5), and then decide to spend 10 minutes searching (active roll at normal DC). What you shouldn't do is let them make an active roll for the walk down the corridor with a PP floor at the same DC - because that's effectively just skewing the probability math with no fictional justification for doing so. There's a metagame decision for the players to make - are we happy to take the average result over time using PP, keeping the game moving and trading the chance of exceptional success for the removal of exceptional failure; or do we want to call out this particular area and go to the dice for the chance to do better than average? What my change does is separate this metagame choice from the completely different choice that exists within the fiction - about spending more time searching in exchange for a greater chance of success.
PS You might also want to check out this post for more on how I tie this into movement and time tracking: http://www.enworld.org/forum/showthread.php?579351-Dungeon-movement-and-Time-Tracking
