There are many anecdotes of hams learning Morse Code at e.g. 5wpm and then "hitting a wall" trying to gradually speed up to 15wpm.
David G. Finley, N1IRZ claims the reason is this:
Morse at 5 wpm and Morse at 15 or 20 wpm are completely different critters, and you don’t want to waste time on the wrong one.
In his essay "So You Want To Learn Morse?" he contrasts the slow "lookup table" of thinking about the individual dits and dahs, versus the faster "reflex" way of learning the whole sounds of letters. If it's not clear, he's in favor of learning the fast/reflex sounds.
I suppose this is a little akin to the Phonics vs. whole language debate in teaching English reading skills, and perhaps there's merits to both sides. But I'm willing to accept N1IRZ's claims regarding the superiority of Koch-style methods for learning fast morse code. Makes sense to me and seems to match my experience so far. What I learn of "slow Morse" won't help me with "fast Morse".
What I'm wondering, though: is the "flip side" true? If I learn "fast Morse" first, will I still need to learn "slow Morse" separately anyway before I can understand the latter?
That is, say I go ahead and stop learning the "slow" whole patterns of letters (so as not to "waste time on the wrong one"). I get good at "fast" letter sounds, and can talk to fast keyers. Will I end up unable to hear what someone using QRS is saying? Or does it end up being easier to hear slow Morse as a stre-e-e-tched out version of known sounds, versus being simply overwhelmed by uncountable dashes and dots coming from the other direction?
Since N1IRZ claims these are "completely different critters", that sort of implies I really need to learn them both to converse fluently with both sets of operators on the bands. But perhaps it is more trivial to understand slow Morse after learning fast?