Generally speaking, a dungeon has enough firepower in it to completely slaughter a party. Generally speaking, a party beats the dungeon. Why does the party win? They have a number of key tactical advantages that players tend to get good at over time. They aren’t usually able to articulate why the things they’re doing work, but they figure out what works and stick to it, leveraging these advantages without even being able to tell other people what they are. So what are they?
First and foremost, concentration of firepower. This is where the axiom “never split the party” comes from. The party’s greatest advantage is that they are all in one group together attacking through the same entrance with the same method at the same time, but monsters have to defend an entire dungeon from a wide variety of potential attacks at all times. This not only concentrates their firepower, it also gives them the ability to communicate with one another very rapidly compared to monsters that need a system of messengers and lookouts to coordinate their defense. The party’s second major advantage is their speed, and it is derived from that communication speed. The party are all in the same room and the party leader can communicate a decision as to what to do next within seconds. They can even have short discussions about how to assault a dungeon in the middle of the assault because they are communicating that much faster than Team Monster.
Team Monster has to send messengers between chokepoints and reserve forces that can be as much as a minute or two apart from one another (especially if their messengers are on foot due to cramped quarters or a lack of wealth to procure mounts). Add that gap in between each sentence in the following dialogue:
East Entrance: HQ, there’s intruders and they’re like level 8. We’re all going to die. Also, Eastern Fallback, activate the traps.
HQ: Reserves, go to Eastern Fallback and get ready to fight.
Reserves: Roger that, HQ. Moving now.
By the time the Reserves get the message, the battle at the eastern fallback position has probably begun, and by the time they begin to arrive, it will probably either be over or at the very least be strongly favoring the attackers. Team Monster is slow and they are spread out because they have to guard multiple entrances, and they need internal patrols and guards so that lone infiltrators (PCs don’t want to be lone infiltrators because they don’t want to split the party, but the dungeon needs to be prepared for all attacks, not just the PCs) will be caught even if they sneak past the heavily defended entrances, and they need fallback positions so that they have a defensible location to commit their reserves to if an entrance is busted open, and they also need defenses on the inner sanctum heavy enough that anyone who punches through the dungeon faster than they can commit reserves or siphon forces from other entrances will hopefully be too weak to overcome the inner sanctum defenses.
But on the other hand, Team Monster has an advantage, too, which is that there’s seriously like thirty of them in even a small dungeon, and they will completely flatten the party if the party is ever forced to fight them all in one go. Team Monster is a vast, unending horde that moves slowly and is spread across a large area. Team Player is a small, concentrated strike team that moves quickly and punches through defenses towards their objective in what is hopefully a fairly straight line. It’s an asymmetric game in which each side tries to use their advantages and shore up their disadvantages better than the other side can do the same.
So what happens if your players completely abandon one of their key advantages to take a rest for eight hours in enemy territory? Well, the 5-10 minute delays between a lookout or alarm bell reporting that there are intruders and the point when reinforcements actually arrive are agonizingly slow when the party is on the move, but paltry in the face of even a short rest. The monsters don’t know it, but they have as much time as they jolly well please to set up a counterattack, and that counterattack will probably hit before the players can complete their rest.
This is not just a roundabout way of saying that if players rest, they should get random encounters. A GM shouldn’t just throw an encounter at over-resting players and then append a verisimilitude explanation post-hoc. A GM should understand how Team Monster fights and use that information to create a living dungeon that reacts to PC intrusion. You shouldn’t just pick up the encounter from the next room over and drop it on them, nor should you have the entire dungeon empty itself out to confront them. Instead, just play the world like it’s a real place. The warlord running Team Monster’s troops in the dungeon (the hobgoblin captain or the necromancer in the catacombs or whatever villain is in charge around here) hears a report from a lookout that the guards on one of the dungeon entrances just got their shit completely wrecked, or else he hears from patrols who show up after the fact and report in that the unit defending a dungeon entrance has been killed to a man. What does the warlord do? He doesn’t have the GM’s omniscient perspective, so there’s several defensible options.
Option One: Panic Like An Idiot.
Look, dungeon warlords are sometimes pretty stupid. Your average orc chief became chief by virtue of being bigger and stronger than everyone else, and is probably of average intelligence at best. All he likely knows is a general location of the party. If he was warned by a lookout who left as soon as it became clear the party had a good chance of beating the entrance defenses, he’ll probably commit reserves to the entrance he’s just heard is under attack without even thinking far enough through to realize it’s probably already fallen, what with the communication delay. If no enemies got away and the warlord was warned when a patrol found dead bodies, his response is probably going to be to bellow “FIND THEM” and commit any off-duty troops to increased patrols. Which means the adventurers are going to get a lot of wandering encounters because the dungeon’s defenses have now been scrambled, but it will be several dead patrols before the warlord starts to put together any kind of precise idea of their location.
What this looks like from the players’ perspective is that they get a few random encounters pulled from other sections of the dungeon, and then the warlord gets a clue and the entire interior contents of the dungeon come down on them. Entrance guards are unlikely to be sacrificed, since even very foolish warlords will probably be smart enough to realize that there could be a second attack coming (even though there isn’t) if they abandon their defenses on their flanks.
Option Two: Cautious Reconnaissance
Let’s say the dungeon warlord understands how his side of the game works and he wants to minimize his weakness. So he’s still going to pull all off-duty soldiers out of bed or the dungeon’s makeshift tavern or wherever they’re being off-duty at, but instead of splitting them into small, easily dispatched groups, he’s going to mass them into a reserve force and commit them to whatever section of the dungeon appears to be next in the party’s warpath based on what areas have stopped reporting in. If it’s not clear exactly where the party will be going next, he’ll want to split his forces to cover the possibilities, and have them be ready to move to reinforce one another as soon as the party attacks one of the groups. Now, the warlord doesn’t know this, but the party is resting. After his forces get in place and it’s been several minutes and there’s still no attack, the warlord will badly want to find them for fear that they’ve either slipped through his defenses somehow or that they only needed to penetrate so far into the dungeon to bring the whole thing down somehow or that for some other reason they have a good reason to have ceased their advance. He’ll send out patrols to find them, and he’ll have those patrols report back room by room. When one of those patrols doesn’t return, he’ll know exactly what room the party are in, he’ll immediately move to surround it (possibly leaving the most direct route to the surface open since forcing the party to retreat is still a victory and will be easier than fighting them to the death), and then he’ll bring the hammer down from as many different directions as possible as simultaneously as possible.
What this looks like is that the party gets one patrol encounter, and a small patrol that the warlord is comfortable sacrificing, and then the interior forces of the dungeon come down on them like the first of an angry god.
Option Three: Counterattack
So let’s say your dungeon warlord is really aggressive, but also has his shit together. He’s going to pull his interior forces into a reserve, and while they’re forming up he’s going to go over the latest reports and try to figure out the party’s objective. Based on their current progress through the dungeon and general knowledge about adventures (they usually want either treasure, glory, or justice), he’ll try to work out where they’re going and move his reserves to intercept them. He’ll have small patrols moving a room or two ahead in all directions, with runners that will inform him if they find the enemy and a half-dozen or so mooks to keep the party busy for the round it takes the runner to report back and the two-ish rounds it will take the horde to arrive. If dungeon layout permits and the warlord particularly wants the intruders to die, he may take a few rounds extra to encircle them.
What this looks like to the players is their rest is interrupted by a patrol, and two to three rounds later the entire contents of the dungeon interior show up to stomp them.