One one more possible result of potential PC death/lethality is the "fight them next time" approach.
Essentially, the PCs engage, then immediately back off if it doesn't look like they can win the fight without cost.
This approach is really common in my games - engage, back off, and come back with tactics, weaponry, magic, and backup plans customized to the specific foe. It's happened against a demon-temple, the bandits in the Cold Fens, the orcs in Felltower, the druagr, and a number of other encounters.
There is a lot of sense in this kind of plan - you never engage willy-nilly, but rather launch deliberate assaults only when you have stacked all of the odds in your favor.
If this always works, though, it's pretty much the GMs fault. Set piece monsters that wait for intruders to come back when they are ready and don't make a second trip a dicier proposition than the first encourage this behavior. And why not? If you're not dealing with a time crunch, you may as well come back when the odds are in your favor.
For a lot of encounters, this makes sense - mindless undead (or area-limited willful undead), guardian statues, traps, ancient temples of the darkest evil, test encounters left in place. They simply can't or won't change. They will stay static and await your approach.
But others will change.
The AD&D DUNGEON MASTERS GUIDE has a great section on this. It's on pp. 105-106, MONSTERS AND ORGANIZATION. It has excellent advice, some great Gygaxian admonishments from on high, and interesting insight into very old school gamers (they attacked a lot of lairs, camps, and towns). It assumes the party engages, inflicts damage, and pulls back because they don't achieve victory.
However sometimes you'll get none of that - just sheer backing off as soon as possible.
I'll try to split these into two general categories - rewarding immediate action, and punishing delayed action.
The Rewards of Boldness
These approaches encourage you to take risk right now.
Opportunity Knocks. Make some encounters have action going on - action that gives a chance you won't get next time. One AD&D adventure has you accidentally interrupt a duel, and the distraction serves to get one duelist wounded - now you have half of your opposition wounded. Other examples are thieves looking over loot, monsters drunk on liquor or drugs (they won't be next time), sleeping guards (who won't be unwary next time), or distracted sentries. Or even just unprepared monsters - maybe they're doing their weekly maintenance on the giant crossbow, or just expended a bunch of spell power to impress each other. You could catch a lot of critters together in one place (a nice target), or catch them strung out (vulnerable to defeat in detail.) Or maybe they're just exhausted - they just finished exterminating some other poor guys and aren't up to another fight.
Make it clear this is temporary - jump them now and you can take advantage. Back off, and next time this isn't there anymore.
Time limited rewards - this approach puts an additional reward in for acting right away. Here I'm thinking treasure, valuable items, etc. that are obviously there . . . but won't last. A rust monster just waddling up to a suit of gleaming and glowing plate armor is a pretty clear choice. So is anything valuable but edible near a hungry monster. Prisoners are a form of time-limited reward. "We'll come back and deal with the mind flayers in a few weeks, and rescue their hapless prisoners then. I'm sure they'll be fine until then!" Yeah, or not. Maybe not.
You can also do a meta-award, if you want to be heavier-handed. Bonus XP for stuff killed in a first encounter, or a penalty for things killed routinely or easily. Rolemaster did something of this sort, actually, for repeat experiences. A bonus for bold action (maybe ala the Awesome bonus) is possible, too. Many players will just weight the cost here, too, though, like it was a sale ("It's +X xp if I do it now, but I might die forever, which is -100% XP and -100% stuff, so it's a bad bargain.")
The Cost of Caution
These basically say, go ahead and back off, but this was as easy as it was going to get.
They just leave - Much like the bandit camp in situation 2 of the DMG p. 105, many foes will just up and leave. They'll take their loot and go elsewhere, expending their resources on foiling pursuit. Flighty opponents who aren't around very often are a form of "Opportunity Knocks." This doesn't have to be monsters fleeing. Maybe they just aren't around very often - the temple that unseals once in a thousand years, or the traveling gold caravan, a temporary moongate, or some other event that is encountered in transit across space or time, not a fixture to be dealt with at leisure.
Reinforcements - Make it clear monsters multiply. They recruit, they grow, they spawn, they go from "small tribe" to "legion of doom" if not dealt with. Some classic AD&D modules did this - T1-4 and WG4 made excellent use of these. I loved both of those adventures, so, so do I. See, the Orcs of Felltower.
The idea here is that monster numbers might be stable, or might make the encounter tougher.
Intelligence - monsters aren't limited to what you show in your arrival and back off. If they have allies, they can get info from them. Or they can us magic to divine more. Maybe they recognize the signs, seals, and symbols. Maybe they've run into different adventurers and prepare based on that experience. But if you come back later, you come back to foes who have prepared for you just as you've prepared for them.
Even if they only react to what you show them directly, the more hit-and-run and show-up-and-run encounters you have with a group, they better they know what to do next time you show up. Maybe even wall you off near a demon lord.
Decreasing Reward - the longer you take, the less reward. The monsters spend some of their money regularly, and don't accumulate it as fast. There is an outstanding reward on the head of one of the critters that someone else might come and claim. Your sponsor gives you a bonus for speedy completion. Basically, the reward will go down. This overlaps a lot with Opportunity Knocks and Time-Limited Rewards.
This puts aside the issue of encounters you can't flee from - room entrances that seal off behind you, attacks on your camp or base, exceptionally good pursuers, etc. That is the kind of stuff that justifies extreme caution, in a lot of ways. That's kind of off-topic here. I'm thinking in the examples of above of giving the choice of caution or boldness, and either rewarding boldness or attaching a cost to caution.
Some of these have happened in my game - some encounters disappeared. Some monsters took their treasure and left. The orcs have gotten so many reinforcements (and money from the PCs, and accidental extermination of their rivals by the PCs) that they have become the preeminent problem for the PCs in Felltower. Others spent treasure on weapons to counter the PCs. Most made tactical adjustments based on what they saw. A few gained intelligence directly and indirectly on the PCs. Still others dealt with their security lapses. And still more made alliances to deal with the threat next time.
Even if backing off is the sensible thing now, or just the tactic of choice, it isn't always without cost. Yet it shouldn't be a universally bad choice, or players will rightly feel railroaded into always fighting before it gets worse or gets away. Hopefully the ideas above will make sure that "when the going gets tough, come back later and deal with it" is not always a successful or wise choice, without making it never a successful or wise choice. You want just the right level of doubt about fight vs. flight.
Here are Part I and Part II of this post series.