On the subject of what is an invalid conundrum...
1. If the conundrum has no valid answer, the conundrum is invalid. (However, if the intended answer is invalid but there's a valid answer, the conundrum stands.)
2a. The answer is a +S plural. These are specifically excluded by the guidelines sent to Countdown contestants, so it's invalid.
2b. The answer is a present tense third person singular verb, like DECIPHERS. I'd advise against using these because they are never used on Countdown (except perhaps in the very early years), and people might reasonably not expect them, but there's no rule banning them. The conundrum is valid.
3. In my opinion, if a conundrum has multiple valid solutions, that shouldn't make the conundrum invalid. On Countdown they may intend every conundrum to have only one valid solution, but on the rare occasions where an alternative solution slips through, they accept either. So any valid solution should be accepted.
If the conundrum is invalid (1 and 2a) then it must be redone, otherwise (2b and 3) it must stand.
On the subject of when an incorrect decision becomes set in stone and can't be changed... this is the more interesting part of the problem, and I don't think co-events have ever had to make any rule about it. Cricket has
law 16.10, which provides that once the umpires and scorers agree the scores at the end of the match, that's final. I'd say the equivalent of this for co-events would be that the result of the tournament is final when the organiser declares the winner.
There are two broad kinds of mistake which I think need different treatment. One is the classic data entry mistake, where the game was played correctly but it got put into the computer wrong, and the other is an adjudication mistake made in the game, which covers words being wrongly accepted/disallowed, or conundrums being misassessed as valid or invalid.
Data entry cockups
There have been multiple cases of scores being entered into the computer incorrectly (usually a reversed result, until I remade the Atropine interface to make this mistake less surprisingly easy to make) and this going unnoticed until several days later. In all the cases I know of, these have been fixed retrospectively and the standings changed. This has never had any significant effect like changing which players should have played in the final, but that's more by luck than anything else.
What if it was discovered two days after an event that there was a data entry cockup which meant 2nd and 3rd were the wrong way round? The "easiest" solution would be to say that any protest about the scores or the positions should have been made at the time, and that's the reason all the scores are displayed on a public screen, but then arguably that should also apply if it was only, say 15th and 16th places at stake rather than 2nd and 3rd. Instinctively it seems better to correct a result you know to be wrong, but if that affects the top two then it isn't even a simple matter of giving the 1st place trophy to someone else - there would have been different players in the final so you won't know who the "real" winner should have been.
A few times I've been aware of a score in (say) round 1 only being corrected midway through round 2, which means the fixtures for round 2 were based on the wrong results, but the only thing you can do here is correct the result and keep the "wrong" fixtures. Especially if everyone's already sat down to play.
So in summary - mistakes that are purely data-entry-related should be corrected as far as reasonably practicable, even if they're reported days after the event. I still don't have an answer to what you do if the mistake would have changed the finalists, though.
In-game cockups
Mistakes made in adjudicating a game are often pretty easy - most of these are covered by the catchall rule in most sports that the referee's decision is final. Anything like who buzzed first, or whether someone's buzz was in time, or whether someone's taking too long to give their numbers solution, are all matters for the host's judgement. People are welcome to bicker about it at the pub afterwards but it won't change anything, except perhaps generate ridicule towards the host in the event of particularly ghastly decisions.
The exception to this is if the referee gets the rules wrong. There's certainly
precedent in football for replaying part of a game - this was because the referee got the rules wrong and denied England a retake of a penalty, rather than because of a wrong judgement call on whether the ball had crossed a line or whatever.
There a number of ways a co-event game host can get the rules wrong - allowing a selection with too many vowels or consonants, incorrectly allowing or disallowing a word, missing a mistake in a player's arithmetic, accepting an invalid conundrum solution or accepting a valid one. I'll include setting an invalid conundrum in this as well, although usually that's not the host's fault.
I Am The Ref
My opinion on how far the statute of limitations should extend for each of these are as follows:
- Host gives you a vowel when you asked for a consonant, or vice versa: raise the problem immediately or it stands. This also applies if the host gives the player the wrong number of large/small numbers.
- Host allows more than five vowels or more than six consonants: you've got until the clock runs out to point out the problem, otherwise the round stands. This also applies if a numbers selection is invalid, for example there are seven numbers in it, or three of them are 6.
- Host incorrectly allows or disallows a word (either because of whether it's in the dictionary or the selection): you've got until the end of the game to raise this, or the host's decision stands. (Yes, even if the selection was EEEMMBEEX and they allowed your opponent to declare FOOTSTOOL - why didn't you point out the problem at the time?) I'm not entirely happy with this because it means someone who knows they've had a word wrongly disallowed in R1 could wait until just before the conundrum and suddenly spring the correct score on their opponent. But I don't have a better answer.
- Host asked the wrong player to declare first: one of you has to raise this there and then before the player gives their word, or it stands.
- One player declares not written down, but the host asks the wrong player for their word first and the NWD player has the same: accept the not written down word.
- Host incorrectly allows or disallows numbers working (arithmetic mistake, not judgement call over whether they took too long): raise this before the end of the game or the host's decision stands. This is a bit trickier than the disallowed/allowed word mistake above, as the player's working isn't always written down by the host so it isn't always possible to check. If after discussion there's any doubt, the host should stick with their original decision.
- Host adds up the score wrong: tricky. When I'm hosting, I usually tell both players what I think the final score is, and check that doesn't disagree with any idea they have of the score. If hosts make a habit of doing that then we can say that any mistakes of this kind have to be pointed out before the scoresheet is handed in. As it is, I'm sure we've corrected score arithmetic mistakes after handing in the scoresheet before. (I think Scrabble and chess have a system where both players sign the scoresheet, after which the result can't be corrected - handing in the scoresheet would seem to be the equivalent for co-events.)
- Invalid conundrum, or valid conundrum answer disallowed, or invalid conundrum answer allowed: raise this before the scoresheet is handed in, or it stands as adjudicated.
To answer the question originally posed, if a player wins the final on an invalid conundrum, and this isn't raised before the scoresheet is handed in, the result should stand. (Next problem: is the scoresheet ever actually "handed in" after a final?)