Here's how I see it.
1E: All about simulation of pulp fantasy novels, and pioneering adventure ideas. "Fafhrd and the Grey Mouser get together with Conan and Gandalf and go down a dungeon." Hey, we can do lost aztec cities. Hey, we can do crashed spacecraft. Hey, here's a chess puzzle and it's not yet a cliche etc. Rules simple in some areas, and fettered with labyrinthine subsystems in others, many of which never get used. Balance between classes mostly an illusion. Encounter design handwaved. Lots of rules handwaved. Unplayable by today's standards. Nice flavour, though.
2E: All about exploring campaign settings. Halcyon days for published campaigns with cool concepts - Dark Sun, Planescape, Birthright, Al Qadim, Ravenloft, even stuff like Living Jungle. Trivial differences to 1E - perhaps the biggest were kits, and the rise of splat. Those kits and splat made the balance issues more apparent than they were in 1E, resulting in 2E becoming the scorned edition. A case of "oops, we forgot the homebrews", as TSR became increasingly irrelevant to people's home campaigns.
3E: All about trying to balance the game, whilst retaining much of the legacy clunk so that people would switch. The rise of crunch-before-flavour, where the monster manual contents were determined with many crunch objectives in mind, flavour as afterthought (a direction turned up to 11 with 4E). Lots of player options in the core, which didn't translate well to stats for NPCs and monsters. But on the whole, faithful to D&D mythology and much more balanced than 1E and 2E.
4E: Create a new game based roughly on the exception-based design of Magic: The Gathering. Remove almost all the legacy clunk in an attempt to make things balanced and fun from a gameplay standpoint (although critically, not from a flavour or suspension of disbelief standpoint, IMO). Try and get the game online. Try and sell miniatures. Try and be hip and different enough flavourwise draw some of the WoW and Harry Potter crowd. Ignore generic pulp fantasy simulation to a degree, but do an ourobouros so that D&D eats it's own tail with an identifiable specific and trademarkable IP in the core and is difficult to relate to anything but itself (a direction started out in OD&D with things like the cleric, but arguably fetishised with Praemal, Eberron, and 4E. Hello dragonborn, hello warlord, hello eladrin). Overhaul flavour so that it's more fun from a metagame standpoint, is compromised by crunch or handwaved when the two conflict, and has a specific, identifiable IP which can no longer do classic pulp fantasy without banning core races and at least one class. All splat becomes core to sell more books.
I suspect 4E is much influenced by M:tG, WoW and GW. M:tG for the exception-based design, random collectables and M:tG Online. WoW and GW for the identifiable IP (Dragonborn vs taurens? Eladrin vs eldar?). GW for firing the customer base periodically - after all, grognards already own an army, so they're of no more use to the company unless those minis are rendered obsolete. I suspect that might have given WOTC the "who cares about retention" direction that 4E has in not tipping it's cap to D&D flavour continuity in the same way that 3E did.