We could speculate that the nacelles indeed move, just a tad, when the ship shifts from Warp 2 to Warp 3 to Warp 4 and so forth. And we fail to see this movement on screen because it's so slight. This "variable warp geometry" could then make some sort of sense.
But it would still defy reason why the nacelles would be built to swing in a giant arc when they only need to move those near-invisible fractions of degree in order to do the job.
It seems more logical to assume that the swinging has got something to do with balancing the ship for planetary landings. However, that, too, is problematic: why do the nacelles swing down for sublight even when the ship is nowhere near planets?
Perhaps it's not the position of the nacelles that is important for adjusting the warp field? Perhaps it's the very movement itself? That is, perhaps the Intrepid class enjoys superior acceleration to warp because it squeezes its nacelles together at the moment of engaging warp. After that initial squeeze, the position of the nacelles is of no relevance - it's only the initial flapping move that matters. And thus of course the nacelles go down when dropping out of warp, so that the ship is immediately prepared for another flap.
Many modern Trek ships have their nacelles at an angle, so such a position may be advantageous overall. Few have flapping pylons, though - in addition to Intrepid, the Yeager kitbash from DS9 looks like it would also have this feature, but that's pretty much it. Perhaps the swing-pylon design was a grievous error, something that never worked the way it was intended? Perhaps it was originally assumed that the nacelles would go up a little bit for Warp 1, a bit more for Warp 4, a lot for Warp 7, and to the upper max for Warp 9.975. It would make some sense, then, to configure the ship so that impulse drive corresponded to the lowermost position. However, this failed to work, and now the decision to have impulse configuration dependent on lowering of pylons turned out to be disastrous. Starfleet couldn't lock the pylons in the single working position now, but had to tell its captains to utilize the two extreme positions, one for all warp, one for impulse.
Timo Saloniemi