As I was writing this mini-series, I've been reading a bit around the internet. Other people also have their issues with Fail Forward, including major "Powered by the Apocalypse!" content writers, apparently.
Something I picked up on in that reading was something that I did touch on myself, but think is a bit secondary in my concerns. That is, basically, is there any connection between what the character attempted, and the consequence of the failure?
The first example has a rogue make a Charisma check in order to befriend an officer on a ship. The rogue fails, and the GM interprets this to mean that rather than fail to befriend the officer, the character did in fact make a good impression, but the officer is now suddenly a cannibal.
Don't get me wrong here. This example here makes me furrow my brow and raise an eyebrow and sort of shake my head in mild confusion; it's just that I'm more worried about other, larger-scale aspects of Fail Forward.
On reflection though, it struck me that this might be down to a bit of a storygamer vs. trad gamer mindset difference. I think the spontaneous cannibalism is weird, as does David Guyll, but that doesn't necessarily mean it's objectively bad. I'm wondering whether this might actually reflect attitudes towards the simulationism or otherwise of the gameworld.
In the kind of simulationist game I run, this example seems silly, because a person is or is not a cannibal. I mean, I might not have determined that in advance (it's not exactly one of the core stats), but equally I sort of expect cannibalism to be a trait I decide on some narrative or logical basis. I would find it weird to have a mechanical roll on a skill change a part of the gameworld reality that isn't even related to it.
To be fair, I might potentially get the idea that an NPC could be a cannibal based on the interactions involved in a skill roll, probably if it made me think it would be ironic in some way. But it'd be a consequence of the narrative rather than the die result.
However, if you don't subscribe to the view that a game is approximately simulating a reality for you in the first place, maybe this makes a lot more sense. Plenty of games and gamers are happy with games where a lot of narrative control rests with the players, and where parts of the game reality are subject to change. To take a classic RPG example, when a players looks for a diamond in a safe, some gamers are happy for a successful roll to mean the diamond is in the safe, without any need for the GM to have decided it was there beforehand.
As far as I'm concerned, this is mostly unsatisfying. But that's just my preference.
Players who enjoy this kind of Conflict Resolution-based approach, and the relatively flexible game reality, may be more amenable to the idea that a game-mechanical consequence of their actions or failures doesn't necessarily match the game-narrative consequences of their actions in a mappable way. If you're tending to view things more as Conflicts than as specific Tasks in the first place, then it may make more sense to you.
To pull things back to our example: maybe for a storygamey player, it's natural to model the Charisma roll as a Conflict like "do I manage to befriend the nice officer", where a failure allows that the officer is secretly a cannibal and so not so nice after all. Maybe Grignr failing to behead an orc and suddenly being endangered by a plummeting chandelier feels like a natural sequence of events simply because it comprises part of a story of sequential events.