Coming back in to kinda try to go back to the OP. For the following I'll posit that "GM Fiat" is when the person placed in the GM authority role makes a decision about outcomes of play that either a) arbitrarily cancels a player's intent and ability (eg: the Alarm scenario in the OP); or b) in the absence of direct mechanics or guidance rules an outcome in an opaque way or one that feels off to the table (this often turns into an argument).
Give above, let's look quickly at principled OSR play - which apart from some very recent NSR style rulesets primarily relies on community norms. It largely avoids arbitrary fiat by use of extensive prep. The players know that the GM is rolling with rigor and honesty, referring to the outcome of tables to allow probability to have its play, and is generous with information to avoid "gotcha" moments. This allows the players to exercise their personal problem solving skills to explore the world and expect predictable results (or in the case of random tables, expect results that they can impact through how long they spend; and what they declare procedurally). Note though: I've only seen this show up in the actual
rules of a game (and even there really just as "you should run this way oh and go read the
Principia Apocrypha") very recently.
Lets now turn to the Blades Position and Effect discussion and mechanics, which I think are by far the clearest way to avoid Fiat (arbitrary and opaque decision making) I've personally run (I'm sure the BW family of games do excellent stuff from what
@pemerton has written about how the players set the stakes of challenges but I've never personally dealt with those).
Position and Effect takes two things that are often traditionally hidden/spontaneous in conventional games and puts it right front and center in the mechanics and conversation. How bad are things going to go for you if this doesn't go so well, and what are you going to get out of it? To determine the first, the GM proposes how they understand the situation (tier, fictional position, etc all inform this) - and the book tells you very clearly "assume Risky/Standard by default, because our actions are interested in Scoundrels facing trouble and they're competent." The GM must then also say "ok, and here's what you're going to
get based on how you're dealing with the risk." From the get go, this is a negotiation, and the player(s) have tons of levers to drive both sides of the equation: they can change their ability (oh, it's a Limited Skirmish here because they're heavily armored? huh didn't realize that - I'm going to toss a grenade instead), aid, trade position for effect (or effect for position, although I rarely see this), pull out items or
fine items, propose or ask for Devil's Bargains, Push (and Push to activate abilities), note playbook abilities, do Set Up actions, & etc.
Essentially, the GM just opens the conversation - and it's not over until the table agrees on what the Risk and Reward for the upcoming dice roll is. And then you roll, and the fiction and actions follow.
Note that there's nothing like, unique about Blades' adjudication system that makes this special. Errant, a D20 game, uses Position and Effect as well to tell you how bad a failure is going to be and what you're going to get on a success. Until that dice is rolled and you commit, nothing has been decided.
