An alternative could be to use miryoku in the qwerty variant and use colemak-dh on the software level, but:
1. I like the idea of connecting my keyboard to any computer or OS and using colemak there right away too much. While I can live without doing the same with my Cyrillic layout, since this is something I use personally mostly, for communication, an idea of connecting my keyboard to someone else's computer to troubleshoot some stuff or code something, I like it too much. Also, I plan to not learn qwerty on this keyboard at all, so that I can differentiate in my head "qwerty on normal keyboards" and "colemak on ergo keyboards" so that muscle memory does not overlap, so yeah, no qwerty there.
2. I'd still have to create custom layouts for myself, I cannot reuse an already made Rulemak because I want my own Cyrillic layout, also I'm not sure if it's made for all OSs.

To sum it up better, miryoku has:

Q W F P B                  J L U Y '
A R S T G                   M N E I O
Z X C D V                   K H , . /
ESC SPC TAB      ENT BCKSPC DEL

(main layer, the last layer is thumbs and is not for scale)

[ 7 8 9 ]
; 4 5 6 =
` 1 2 3 \
. 0 -

(num layer; now that I think about it, the duplication of dot on thumb layer is kinda weird)

{ & * ( }
: $ % ^ +
~ ! @ # |
( ) _

(sym layer — now dot is replaced by a duplicate of open bracket for comfort; still kinda weird)

Here's what Ukromak will have:

Я Є Ф П Б    Й Л У І Ю
А Р С Т Г    М Н Е И О
Ц Х З Д В    К Ч , . Ь

(alpha layer; thumbs don't change)

[ 7 8 9 ]
; 4 5 6 =
ї 1 2 3 ж
. 0 -

(num layer)

{ ? * ( }
: ' % ^ +
Ї ! " # Ж
( ) _

(sym layer)

And now the alt layer that is independent from miryoku (on latin alt layer I'll probably have some international characters including Portuguese):

Я Э* Ф П Б       Й Л У Ы* Ю
А Р Ш* Щ* Ґ*    М Н Ё* И О
Ц Х З   Д В       К Ч , . Ъ*

(alpha + alt layer; where letters aren't replaced I could duplicate the ones from the main layer to prevent errors during fast typing)

Yes, I'm mostly happy with it. Only one consideration is that it feels weird to have some letters on num/sym layers, some on alt layer. Maybe I should just bring everything to the alt layer since I'm already using it. So Ї could be alt + И, and Ж could be alt + Р (or П, Д — not sure what's more ergonomic, home ring or ring upward/downward movement; probably ring)

________

K is 3 times more common than Ч (3.486% vs 1.45%) but it's not like Ч is not used at all (e.g. Ц is just 0.486%) so I'll probably keep it the way it is to make it phonetically similar. After all, K is one of those rare letters that look the same and sound the same in Latin & Cyrillic. Only other such letters are... M & some vowels?

Could you sum up without referring to Miryoku, what changes you propose from Rulemak to Ukromak? And reasons therefore based on usage frequencies and n-grams.

I'm planning to add Ukrainian Colemak to EPKL soon – should've done it already, but for time constraints.

Keep in mind that I'm trying to implement a phonetic layout to keep it simpler for users of Colemak, so it won't be optimal for typing Ukrainian just like Rulemak isn't for Russian. At the same time, it shouldn't be horrible either so as long as we can do so we want to adapt. This seems to be what you're doing too.

Also, since this will be a layout for people who type both English and Ukrainian it will be desirable to keep things that are only or mostly for Latin script (such as the <> you mention). There will be times when a user needs to type some Latin script in between the Kyrillic.

Colemak-eD-Kyr in EPKL

Yes, of course! First I mapped everything from Russian to Ukrainian the way it's done in JCUKEN even if it's phonetically weird (e.g. ы -> і, и -> и, even though phonetically it would make more sense to make it неіо instead of неио).
I also replaced / and ' to put more characters on the primary layer without resorting to the alt. To combat this, I put " on shift + 2 and ? on shift + 7 (because this is how it's done in jcuken and the replace @ & that are not used in cyrillic at all), I also put ' on shift + 6 but it was a subjective decision.
I also swapped З and Ц because З is much more frequent, in partcular a ЗА bigram, but understandable if you want to keep it more phonetical.
Then I put Russian letters on alt layer the way it's done in macOS Ukrainian layout, so і -> ы, є -> э, ї -> ъ. Oh and also ґ is alt + г of course.
I think this is it. Also put lots of letters on alt layers but that's because of keyboard limitations.

My take is probably to prefer coming from Colemak (favoring неіо as you say). But there's also the matter of similarity to Rulemak which is a minor benefit to also be considered, so I think I'll agree with you on keeping the неио from Rulemak/JCUKEN. I'm not a fan of keeping multiple versions around just because, as it causes confusion and decision paralysis, as well as tons of extra work for (an) already overworked implementor (implementors).

If we settle for one solution, it should be somewhat open to testing and tweaking by people who use it, but at some point one solution will just have to be considered good enough and universal enough. Anyone wishing to tweak that further are of course free to do so, but unless their variant really catches wind and becomes very popular it'll have to be their personal variant and not an "official" one.

Oh, I now remember that it was you continuing the Rulemak Forum thread in 2020. Nice! But to recap and then move on then, here's what I'll wish for now. This is the current Rulemak, again:

Rulemak (2016):
===============
ё 1 2 3 4 5 6 7 8 9 0 - ъ
    я ж ф п г й л у ы ю ш щ
     а р с т д ч н е и о ь э
    ´ з х ц в б к м , . /

    Rulemak Russian Kyrillic layout, by ghen (2016 version)

Selecting Colemak-Ru_ISO_CAW in EPKL gives this:

Could you make a diagram like this one, for Ukromak? You know, I don't speak the language and I get a bit confused by all the suggestions and stuff. Hehe.

ё 1 2 3 4 5 6 7 8 9 0 - ъ
    я ж ф п г й л у ы ю ш щ
     а р с т д ч н е и о ь э
    ´ з х ц в б к м , . /

@ghen ъ is not used in Ukrainian so ї could replace ъ, but I don't know, maybe ъ is already on alt-layer in rulemak? Then putting ї on alt layer also makes sense.

Other than that, spot on! Though I'd prefer the Ukromak name, just like Bulemak has its own name etc. Up to you of course, I'm personally fine with it, just that there are a lot of traumatized Ukrainians now that get triggered by being associated with anything Russian. Not to throw shade on the Russian language, that's just how things are, unfortunately. Sorry for bringing in politics :/

Also thanks for sharing your layout, maybe I should also put Ж on W, instead of Є/Э. But I suppose it's justified in my case as it's a dual-use key and I simply don't have more keys to put Є on. For Ж in my case, home index on alt layer isn't that bad of a position after all, and in line with colemak's "index fingers are for consonants".

Some suggestions, as discussed with DreymaR in the Colemak Discord.
• In order to add missing characters, map `~ to ʼ₴, as in the Windows Ukrainian layout. The apostrophe can be either ASCII (U+0027) or the more typographically correct modifier letter (U+02BC) according to personal preference. I personally prefer the latter, as it’s proper writing and the ASCII apostrophe may be accessible by other means.
• Move їЇ to =+. It’s still around 0.8% in frequency, so it deserves a better spot than `~.
• Relegate ґҐ to AltGr+7. It’s vanishingly rare enough (0.02%) to be put off the main layer, as is done in the Windows layout for ANSI keyboards.

The ASCII single and double quotes are included in the current AltGr with the Latin script, I think. Gives you a choice between ` for a ‘proper’ apostrophe and AltGr+ь for an ASCII apostrophe.

If we’re reassessing base Rulemak, then I’d definitely adopt a Kharlamak-esque AltGr and dead keys. There’s a Latin–Greek dead key in there too, for scientific notation mostly. For longer texts you’re better off just switching layouts IMO. In nearly three years of testing I haven’t used the dead key as often as I’d anticipated anyway.

I could definitely implement that in EPKL. I’m also working on a huge Cyrillic add-on for EPKL, incorporating multiple layouts and Compose sequences. I’m finally satisfied with the design and symbol coverage, so the new Rulemak implementation may be a teaser for the storm that is to come :-)