When an NPC that was let free to run off comes running back with reinforcements... the quantity and power of those reinforcements is determined many different ways based upon individual DMs. For some DMs (like pemerton it seems)... the decision on the reinforcements comes down to what will make this new fight interesting in and of itself. And the monster group that shows up is based almost entirely on creating that interesting fight (based partly on the monsters further in the dungeon that might show up). For other DMs (like Elf Witch it seems)... the decision on the reinforcements comes down to who that NPC might logically have come into contact with, and convinced to join the fight.
As a DM I had a situation where an encounter turned out much harder than I had planned and the party got their butts handed to them. I had been rolling listening checks to see if the goblins in the guard room heard them. When it went so bad I stopped rolling and decided that the guards were drinking and making to much noise to hear. The players did not know this so they hurriedly searched the room, put a guard on hall and then hide the bodies while the wizard cast prestidigitation to clean up the blood. Then they went in search of a place to hide and rest. They were low on any kind of healing. I made a snap decision that one of the scrolls was rope trick so they could safely rest without worrying about being interrupted.
For my part, I don't see the radical difference between deciding that an escaped NPC hides in a cupboard, and decding that the PCs discover a Rope Trick scroll so that they can take a safe rest. They both look to me like GMing devices to regulate the pacing of play.
I don't run a Gygaxian D&D game - I've tried and I'm not a good GM for that style - but I think I have a pretty reasonable understanding of it. The dungeon is keyed in advance, and the GM doesn't depart from that key. This is what then enables the players to exhibit skill in play, by scouting and divining and so on in the way that Gygax describes in his PHB, and then settling on targets and launching dedicated raids for planned purposes. If you are playing in that style, then the behaviour of the escaped NPC should be determined by a combination of pre-planned facts about the personalities in the adventure (which the players can learn about via rumours, ESP etc) and a dispassionate application of the moral rules.
But once you are not playing in Gygaxian style - and once you start placing Rope Trick scrolls out of sympathy to the players I don't think you are - then I personally don't see much point in maintaing certain of its tropes in an illusory fashion. If the unexplored parts of the dungeon are morphable then they are morphable - not just in terms of the rope trick scrolls they might contain, or the drunkenness of carousing goblin guards, but in whether escaped NPCs will bring back reinforcements or hide in cupboards, and also in whether those reinforcements will take 3 minutes or 5 minutes to turn up (and for a 4e short rest we're only talking about 5 minutes), and even in whether those reinforcements are goblins warriors (standard creatures) or goblin cutters (minions).
this kind of thing can't be solved by a single ruleset. Not in the least. We HAVE to have options for both sides available to us, because I don't think there's even a close-to-consensus on one side or another.
Whereas I don't think it's primarily a rules issue at all. It's about (i) the purpose of play (Gygaxian skilled play vs someting else), and (ii) about what sort of approach to encounter design will provide a fun play experience relative to (i) and also relative to the action resolution mechanics.
Assuming the designers are clear about what their action resolution mechanics are (eg is the game fun even if players don't have access to their short-rest recharging abilities?) then they should be able to write play advice that canvasses a range of answers to (i) and hence a range of answers to (ii). If they want to, they can even canvass a range of possible short rest durations and explain how choices in respect of that will matter to both (i) and (ii).
This is hardly a trivial dial. While DMs can easily change the resting time for their tables, published adventures depend on knowing how long the rest is, as they have to be designed totally different when the PCs can rest between battles for 5 minutes and get their power back or not.
The difference between 5 minutes and 1 hour is significant in a dungeon environment, for the sorts of reasons [MENTION=18333]Neechen[/MENTION] mentions upthread. But once you are playing adventures set in countryside or urban areas it's actually not that significant a difference.
As far as published adventures are concerned, I would expect the designers to explain at what point they expect the PCs to be "recharged" before pressing on or encountering further enemies, and to give some advice on how this might be handled at different tables using different resting times. This is hardly rocket science in adventure design.