Simple--that's when you put the focus of the game back to the paladin, and give him a chance to escape. If he makes it out, extra XP from the harrowing tale in tow, then he goes back to wherever and forms a new party.
By the way, how does the paladin's player feel about all this? Is he okay with everyone running away?
I like that from the fairness angle, and it's pretty close to what I actually did. I invented at the table, a
klah-tasting potion for the orcs to stabilize the paladin, which felt like distasteful fudging, and then ran a split adventure where the paladin and warmage made a deal with the slaves' bookkeeper and then fought their way out with the aid of the slave ogre who powered the elevator, madly leveling down encounters as I went.
However, even if the paladin were the focus of the new party instead of the flee-ers, it retains one of the things that makes me avoid PC death in the first place: the whole "no more adventure for two hours, we need to go get new characters built/raised/healed/integrated into the party."
What do the players want? Do your players enjoy having the life or death of their characters decided by dramatic fiat or do they want such decisions to be more in their control?
By deciding dramatically when the possibility for death exists, you are removing the consequences from any tactical choices the party cares to make. This is something the players really should be on board with if you plan to do it.
Nope, the players all seem not to mind death and sometimes beg for consequences. (I do a lot of consequences as far as "this happens as a result of your doing this," I just don't do a lot of "you did this and so I'm going to nail you to the wall.") I just can't rid myself of the feeling that "D&D" is when you have a group of four characters opening the door to a room and killing a monster, and when you are ripping up your character sheet after dying in a random encounter or splitting up the party because your character isn't motivated to go along, you are not playing the game any more. I'm not sure I can find common ground with my players on this.
Completely agree on this one, unless the entire area has already been explored and mapped by the PCs.
I don't reveal the entire map, but the entire map is on the easel pad so you can basically use your intuitions to figure out if things connect and how far a trip it is. I believe in both of these encounters the path was 90% known.
The Secret of Smuggler's Cove is a very nonlinear adventure and the PCs had seen all kinds of connecting rooms without maybe going into every door. In the paladin encounter, the PCs had worked in the room as slaves but didn't know there were sentries to sneak around.
If the flank is achieved in 1-2 rounds, then it might be okay--inefficient, but okay. If the flank should take 1-2 rounds but somehow ends up being 3-5 rounds, or when the combat is almost won, then there's a problem.
I think my problem is encounters that will kill a character in two rounds; if they were lower level, this wouldn't be as bad. The scorpion was actually two levels higher than the party; I was experimenting with sandboxing by scattering modules all over the map, but I wasn't happy at all when I wasn't able to "let the players go with a warning" because they couldn't win and the scorpion couldn't let them go. If they had been using efficient tactics, they maybe, I hope, could have gotten through that one encounter and realized they didn't belong there. With the paladin encounter, they had one extra party member but two extra orcs compared to how the module was written. In contrast, in the combat where the one character didn't fight because he didn't have enough evidence and another never came inside, the combat wasn't rigidly planned out so I just made all the minions 1d4+1 hp commoners with no armor and it went OK -- I just wanted to stall combat a little longer to try to get more people on board.