I wonder how much of the shift in the expected power level of D&D characters over the years is a result of the game's not having an integral mass combat system with which to provide additional context. Consider: OD&D arose out of the fantasy supplement for Chainmail. The default combat system for OD&D is the Chainmail combat system. If you're using that system, you can see directly how effective a Fighting Man is supposed to be against masses of armed and armored troops.
The problem is that, by all accounts, most players of OD&D didn't use Chainmail, but instead chose the "Alternative Combat System" that we all now associate with D&D. That system provides a few hints about the level of "realism" it's supposed to model -- such as the Fighting Man's multiple attacks against creature of 1 Hit Die or lower -- but it's much more "abstract," since it wasn't intended to work in conjunction with larger numbers of massed troops. Gygax attempted to rectify this with Swords & Spells in 1976, but it comes across as more of a "pure" miniatures wargame that's compatible with OD&D than as an adjunct to the game itself. AD&D had no mass combat rules at all, till 1985's Battle System, but I suspect it was too little, too late. Like Swords & Spells before it, Battle System was an adjunct to the rules and most players saw no need for it. 3e had no mass combat rules for its entire run and I've seen no evidence that 4e will have one either, but the game is still new enough that anything is possible.
I've long argued that an important part of understanding how OD&D was supposed to feel lies in understanding its endgame -- stronghold construction and domain management -- but my argument is difficult, because D&D, with the exception of the Mentzer boxed sets, has never adequately discussed these topics. Likewise, as I noted earlier, the lack of such rules has also contributed to power creep by not providing context for the rise in level of characters and their relationship to the wider game world. What this suggests to me is that OD&D desperately needs to rectify this situation in some way, although exactly how I'm not certain.