The DM's Guide to DMing

Some alternate ways to deal with the everyone-rolling-until-we-succeed issue:

Burning Wheel both requires you to define the consequences for failure before the roll, and makes the roll stand until circumstances change. If there's not an interesting consequence to the roll, the DM most likely should just let the player succeed without rolling. If the player does roll, success or failure, the result stands until something significant changes.

Dungeon World asks you to, first, describe how your character goes about doing something before the move is triggered, and, second, failure lets the DM make a move. Which, in Powered by the Apocalypse terms, basically means that something bad happens as a result. Even a partial success can snowball into trouble, so rather than a failed roll being a whiff, it's treated as a "No, but..." Everyone can roll, but they each risk consequences from each roll.

Fate uses consequences and fate points. It's one reason why I like the Dresden Files RPG, because it closely mirrors Jim Butcher's writing style. This isn't quite the same as the original question, since it depends on the fate point economy and doesn't directly link to failed rolls; Fate Core mostly deals with it by only rolling for interesting failures, though their suggestions for how to handle a failed roll are useful and broadly applicable.

While they have some assumptions that make porting them verbatim to D&D have side effects, the principles can still be applied, depending on what results you want. Either let the consequences of the first player's roll stand, or have each individual roll risk danger.

Note that these systems tend to prefer consequences that fail forward (i.e. keep the plot moving) either by giving them the result with a twist (Yes, but...) or introducing another complication (No, but...). They also tend to give experience when characters fail (in Burning Wheel and Dungeon World, but notably not in Apocalypse World or Fate).

I like both of the other solutions too, depending on whether you want to prioritize: increasing risk becomes a push-your-luck game for more dramatic swings, whereas average-the-rolls gives the party a choice between playing it safe for a middling result or relying on the expert for a hopefully-good result. You still have to deal with the occasions where the entire party needs something and blows all of their rolls, but that's a separate problem.

Also, if you're going to house-rule it as a DM, I'd recommend picking one approach and sticking to it, so your players know what to expect.

I usually adlib degrees of failure or success depending on rolls with my 5E games and my players seem to enjoy it. I've come up with some pretty fun side effects from crit fails. I usually try to present the players with situations that don't undermine their task entirely but give them another obstacle that they can surmount.

Probably a good thread to mention The Unmarked, an RPG we're recording and releasing for the next while. Rob Daviau is our DM and I won't say he's exactly making it up as he goes along, but he's VERY light on his feet and nearly seamless. It's pretty fun as a player!

I've yet to have a player character die on me, so this is a bit academic. (NPCs? Sure, go ahead and stab my main villain in the face during his monologue.)

Dungeon World explicitly does it a bit like you just did: when a character runs out of hit points, they get to roll "Last Breath". On a partial success, they get to make a deal with Death.

As a general rule of thumb, I try to limit my taking narrative authority away from players to the means that are carefully codified in the rules. At one point, this lead to a player character becoming a villain and a GM NPC (at the player's request). Lots of indie/hippie RPGs mostly hang out in this space.

D&D has a specific set of assumptions about death, including the availability of resurrection spells. How you choose to work with that is, I think, the biggest lever you have to adjust to the circumstances. There's a bunch of options, but mechanically it mostly works. The main drag is the fictional difficulty, not the mechanical difficulty. Any downtime longer than the rest of the combat will tend to drag a bit. There's tricks to deal with this: let the player play another character (either a new one, or an existing follower or acquaintance) or just skip forward to dealing with the resurrection.

(Mouseguard has one of my favorite fictional dodges for handling both players not being present for a session and bringing new characters in: because they've all sworn oaths to the Guard, they can be ordered around between sessions to invent excuses for the team composition.)

At the total other end of the lethality scale is Old School-style D&D, which is less literally Old School and more Gary Gygax roguelike lethality. Characters will die and you probably won't be able to afford to get them back. If players are on board with this, it plays up the tactical aspects of original D&D. (Paranoia parodies this, with the clones and the killer DM Friend Computer.)

Related to this is Burning Wheel, which is technically less lethal because characters are more afraid of dying: getting wounded will usually knock you out of a fight before you die outright. Wounds put a serious cramp on your style. (I had a dwarf once who lost his hand while adventuring. He finally got stabbed with a morgul blade after taking down a Nazgul single-handedly, and had a lingering death. It was an awesome way to go: he fulfilled his beliefs.)

It’s tricky. If players assume that they’re expected to take care of a problem as soon it comes to their attention, which is a pretty reasonable assumption given the way modules tend to be structured, then it’s hard to blame them for running off. Of course, that’s cold comfort when everyone’s dead in a TPK because they ran off and got themselves killed in response to a bit of world building you were doing.

If you can, try to foreshadow some of the dangers the PCs should have to expect to deal with, but if you have any concerns that the players might not understand, then don’t hesitate to be explicit that they won’t be able to deal with it as they are now (e.g., they need to get more powerful, find allies, etc). You could also try being mindful of the information you reveal, only giving them enough once you’re fairly certain they’ll be able to deal with the problem. However, there’s a risk they’ll figure it out too soon and then go off before they’re ready.

Giving them enough tasks that they have to prioritize isn’t a bad idea, but it can lead to analysis paralysis depending on the group. However, I do like TheHarpoMarxist’s suggestion to give them diversions on the way to the mistake. This is especially good if the reward for completing those tasks is something they can use to their advantage completing the original task.

A couple of techniques to control this I've used is to have an NPC give feedback that they're not ready yet or to 'gate' the access with something that the PCs can't do yet eg a higher level spell or a social status they haven't achieved yet "only a priest of the ninth circle can open the portal"

However since this is usually intended to be a foreshadowing technique my preferred approach is to give them only part of plot hook so that they can't proceed anyway. "The book says there's an entrance in the dungeon and the magical password is.... *turns page*... Oh dear, the next page has been torn out". Then the players can be redirected to the next adventure if they want to pursue the hook. "I hear the library at Castle Crag has a copy of the book. Maybe we should take that job protecting the merchant caravan that's heading that way."

That’s a pretty awesome technique, which I love using too. Just because it was mentioned earlier, I’d like to point out that it’s also one of the GM’s principles (“ask questions and use the answers”) in Dungeon World. Even if one never gets a chance to run or play it, it’s worth reading just for the stuff on GMing. They’re things GMs were already doing, but perhaps not mindfully. I also like the concept of fronts, but it does an amazingly bad job explaining them. It took me a little while to really grok them.

Yeah, that’s definitely a lot like fronts. However, grim portents are not sequences of events but lists of GM moves that one makes in order to bring about the danger’s impending doom. It’s mentioned in an almost throwaway comment, so it’s easy to miss. I haven’t read Apocalypse World, but from what I’ve seen of AW2E, it’s much more explicit. The idea is that when you make a GM move, you can turn to your active dangers and use one of the grim portents as a move. This can be particularly helpful when adjudicating moves like Undertake a Perilous Journey. Sure, you could have them get lost on a miss, but a miss is not failure—it’s a golden opportunity for the GM to make a move; and getting lost is boring anyway. Instead you could look at your dangers, see a grim portent like “The slavers come ashore”, and tell the PCs: “As you arrive back in town from your journey, you notice that several boats have been launched from that ship with the yellow sales in the harbor—you know, that one you identified a few days ago as belonging to slavers. You see a sizable group of them ahead, going door to door, grabbing anyone they can, using just enough force to make their cargo pliant but not dead. They’re all carrying cutlasses but a few also have flintlock pistols. What do you do?”

Right now I’m running a fantasy Fate Core hack, so the power levels are all kind of flat, but I generally prefer to base things on what makes sense in the world rather than what balances better. However, we’re going to be wrapping up in the next few sessions. After that, we’re going back to Pathfinder, but I’ll still be using fronts, which I’m using in my Fate Core game as well (the example grim portent coming from one of my dangers). We’re not doing Dungeon World because the conflict resolution mechanic was just a little too disassociated for some of my players, but I’m going to be borrowing elements we did like (such as answering the end of session questions for XP).