Though they seem generally decent (at first glance), the bizarre thing I noticed about these colemak optimizations is that they seem to put quite a lot of same-finger on the right index finger.

In the first layout, "lo" and "ho" are pretty common digraphs.  I also see no improvement (in fact, I'd argue deprovement) with the "you".

With the other optimization:

I would probably (at the very least) swap g -> v -> q -> g, p <-> j.  Those top row positions are (in my opinion) annoying and not worth putting even slightly common letters on.

Even then, however, I'd still find the right index same-finger glaringly weird.  "ha" is a very common digraph, and "pa" is pretty common as well.  True, the swapped positions are pretty good for alternative fingering, but that's a lot of alternative fingering.

If you'd like, you can check out my experimental layout [optional: unwiden, moving the z and q back to their original positions, swap m,g,d depending on preference, change punctuation if needed].

A couple of things (in addition to the details in my main thread), though:

• The layout is experimental, a work in progress, etc.

• NB: my touchtyping technique (as shown in the diagram) might not be the same as yours'.  This could affect the results.

• "ea" is a common digraph, though not as common as the first layout's "er".

• "oe"/"eo" is same-finger, since it's a reasonably uncommon digraph.

• "you" can be kept as is, or "y" can be swapped with "f". "f" is a lot more common than "y" in digraphs with "o" and "u", however.

I would be a lot more conservative in the second transition.  Relatively speaking, there's no longer much to gain and a lot to lose.  To this end, my current practice regimen prioritizes retaining colemak fluency. Practice was done via "reading" in amphetype, chatting.  Colemak was always practiced immediately before and after, and used exclusively as the daily driver.

At about a week or two, and through significant practice in both, I have so far have managed to retain both layouts, with about 50wpm on the new one. However, it's getting harder to switch back to colemak, with more time spent typing gibberish before I "remember" the layout, so I'm ultimately not sure if this strategy will be successful.

As for "how long", I personally plan to get to the same speed so as to compare the layouts with similar fluency.  This strategy is a bit slow, however; I anticipate it will be a matter of weeks at minimum.