The determination is neutral as long as it does not incorporate bias on the part of the DM. The DM doesn't care whether the PCs make it there, or not. There is no incentive for the DM to contrive anything that is not in his or her honest interpretation of what should be there. The party probably won't get stuck in the market unless there's a festival going on. They may encounter thugs that were specifically sent to detain them, but otherwise, random bandit encounters usually are covered by existing tables (if the DM is uncertain). They probably won't bump into an old acquaintance in the next hour, given the relatively narrow window of opportunity for such an event, unless there's a good reason that there would be an increased likelihood that a particular acquaintance might actually be there.
I can't quite get my head around how you could come to this conclusion. Off the top of my head, I can think of several cognitive biases that constantly assail every person during play (including the GM), the GM during prep, and during post-game evaluation:
Availability heuristic bias
Curse of knowledge bias (probably the biggest problem that GMs in such a scenario face)
Framing effect bias
Observer expectancy effect bias
Selective perception bias
You can only on rare occasions utterly "defeat" these biases or, for that matter, even be aware that your objectivity is being assaulted by them (even if you're a person possessed of hyperawareness). This isn't even bringing in various mundane features of everyday life (lack of sleep, distraction, general fatigue, other stress).
Further, I believe that you cut your teeth on AD&D2e? That game is much more exploration centered than 3.x (which I believe that you also play/played). The generic process for a GM prepping an AD&D exploration scenario involves:
* Establishing what locale/terrain type this specific section of the adventure will take place in.
* Establishing
what is concretely
where.
* Establishing the very specifics of the
what.
* Establishing movement rate per exploration turn in a specific locale.
* Establishing frequency/% of random encounters with respect to turns.
* Establishing encounter tables; populating them with kind/types, lethality, % chance for each.
Merely the last two, encounter table construction/management, are utterly rife with various biases. Further, none of the above even involve the myriad offscreen/undiscovered backstory evaluations/references (temporally and spatially) and the various rulings that you're going to have to make throughout the course of a session (because various subsystems/PC build mechanics are utterly incoherent with respect to each other).
Even if you've (meaning people running a game - not you personally) managed to convince yourself that you're running a cognitively unbiased, perfectly parameterized model run of a high fantasy world in an AD&D game...while simultaneously not falling prey to any of the above cognitive biases when actually running the game (eg having a conversation with your players), consider the inescapable reality that each and every one of your players at the table (who are trying, and failing to one degree or another, to manage their own cognitive biases) are (mis) perceiving and assimilating the (mis) information they believe you are conveying from moment to moment via their own biased mental models.