Speed is instantaneous. It's just one-dimensional index of refraction. When light passes through glass, its wavelength (as measured in a given medium) doesn't change just because it spent some time* in a medium with a different velocity.
(*Since photons have mass = 0 and velocity = c, they can't even experience time. From that frame of reference, generation and absorption happen in the same instant.
)
You shouldn't need to calculate your feed line, but if you are using lengths of coax for matching, yes, the physical length differs -- by the ratio of velocity factors. We generalize this by speaking in terms of electrical lengths: 1/4 wave might be 1m in coax, 1.5m in space, 1.2m in twisted pair, 1.4m in ladder line... or 10mm in a chunk of ferrite. But regardless of the physical length, the electrical length -- some proportion of the wavelength, in whatever medium is being used -- is all that's important.
Antenna designs aren't much of a basis for comparison. They suffer worse than appnotes -- the latter at least has to have some supporting data, and has the tacit approval of a big manufacturer. Antennas can be literally a spool of wire and a pamphlet tossed into a box. Almost no one bothers to measure them, and opinions consist more of superstition than science.
There are certainly practitioners of the craft who adhere to the scientific method. If you see measured (not simulated -- but that, too, is at least a start) data, including feedline impedance or SWR, and radiation pattern and/or gain, you have found a trustworthy source.
The other catch is tuning: even a very poorly cut antenna can still be tuned to a desired frequency. Sometimes depending on feed line length (which, length alone won't match everything -- you need two sections of different impedances and lengths to pull that off -- but it can convert an odd reactance into a resistance, or anywhere else around the circle of phase shifts), mostly depending on a tuner box (usually an LC or CLC network, something like that, with adjustable/selectable values).
And keep an eye out for unbalanced antennas. There are a number of inordinately popular designs out there, that -- probably intentionally -- send signal current up the feedline. They often make claims like, higher bandwidth or directivity than the overall dimensions of the antenna itself would suggest -- or can even permit** -- and customers often find inconsistent results (gee, I wonder why!) or dangerous results (large transmitter feedline currents, anyone?)!
**There is a physical limit relating gain * directivity * bandwidth, and the smallest enclosing sphere the antenna can fit within.
Anyway, back to theory: any odd multiple of 1/4 (i.e., 1/4, 3/4, 5/4 = 1 1/4, etc.) works as a whip antenna (giving an additional lobe in the radiation pattern as you go up for each mode), and any odd multiple of 1/2 (total length) works as a dipole.
Radiation is prevented (by interference) at even multiples -- though not totally, as the radiation resistance doesn't go perfectly to zero or infinity, but remains finite. Mind you, the fact that it's very different from nominal impedance, means it's going to be much lossier to get there in the first place. That is, most of the energy is trapped as reactive power, gobbling up losses but not accomplishing anything.
Or, I would arguably say that radiation does in fact go to zero***, except that no real antenna is perfectly balanced. If you attempt to tune up a dipole at 1λ, what you're matching is not the radiating mode you thought, but something due to asymmetry in the antenna itself: the driven frequency not actually being perfectly harmonic, but differing a little from the antenna's natural antiresonance (and thus still having some radiation left to give); unequal lengths, or bends or kinks or bowing, in the elements; unequal proximity to nearby structures; imbalance in the balun, particularly feedline currents from a shitty type of balun like a current choke; etc.
(***Unless I'm forgetting that a dipole at 1λ actually radiates just fine with axial lobes or something. I don't think so, though?)
Tim