I dunno, I guess my players know that I'm making it all up? The one thing I haven't made up is the result of the dice roll which is why that seems sacrosanct while the rest is up for editing before or during the session.
I also oppose fudging monster HP, AC, reinforcements, etc. once battle is joined. If you give a
reason, if you make it something the players can
learn, then it isn't fudging, regardless of when you do it--modify to your heart's content so long as players are able to learn it. If it's secret,
that's when there need to be strong limits.
To follow on from that - is it OK for a DM to come up with a new strategy during a session despite them devising an initial strategy before the session because the players did something unexpected?
I don't really know if I can say, because I don't know what "a new strategy" means. Different tactics, as in a person making a different tactical decision? People change their minds all the time, that's not strictly a problem. Radically different personality and priorities, such that you're effectively deleting the person they were and replacing them with a clone that doesn't act the same way? That's probably a bridge too far.
I guess I wonder where the DM license to be creative with the fiction during a session starts and ends?
I don't really understand what's confusing about it. Has the thing entered play? Then it should not be altered unless the players can* learn about it. Has it not entered play? Then there are few limits (not
zero limits, rationality remains a thing, but the limits are otherwise very few.) Hence my practical examples above, where the players had actually
researched the type of slime they mean to fight, or examined attack sites and interviewed survivors etc., and thus learned these slimes are weak to cold. That weakness has entered play, because they've actually engaged with it. Other aspects of the slimes might remain unknown--do they attack as a group, or as a single monster? How intelligent are they? Can they fission into two new creatures if damaged? Etc. All those questions could remain entirely unanswered, and thus completely free to alter up until the moment the party starts having an answer for them.
If you want some kind of rule of thumb:
Once the party actually
sees the creature, don't change its stats unless you justify it and give the players a chance to find out what happened.
If the party hasn't
seen the creature, but has done work to learn about the creature, anything they have learn is durable. You must justify any changes, and give the players a chance to find out what changed.
Same goes for things like "who killed the Countess," "the true identity of <player character's> father," and other factual/situation things that could be investigated.
Or if you want a
really pithy maxim: "Don't contradict what the players have
started to learn unless you justify it & let them learn anew."
If you DO justify it and let the players learn anew, then you by definition aren't fudging (it isn't
secret, so it can't be
fudging), and thus no problem occurs.
*It's not that they
must learn of it--just that they
can learn of it. But it must be a genuine, serious "can learn," not a fig-leaf excuse. Again, having to roll a double crit to learn it isn't being sincere about enabling players to learn, but you also don't need to shout it at them either.
---
To turn a question back on you: Why is it you seem to think that an opposition to fudging means putting a collar on GM creativity? Surely deceptive practices are far from the only tool in the GM toolbox. Indeed, I would hope that such a tool (as nearly everyone agrees, even
very pro-fudging folks) would be used very sparingly if it is used at all. And if it
is used so rarely, then opposing it can't be that much of an imposition on the creative GM. This isn't telling someone to paint without using any form of blue; it's telling someone not to use fluorescent yellow-green, a color that would rarely be needed and where its absence can be worked around without losing much of anything.