I'll be honest, I
don't see a need for "theme protection". Mostly because it can't happen in the first place, and that it doesn't actually happen in the second place.
The roles that actually exist in D&D are the same that exist in any conflict; and thus war games, and examinations of war needs to be studied for the roles that it demands of its participants.
Since any sort of "binary results" system is little different from
war game, and the major concerns that we are having have to do with what the character can bring to their war party; that fundamental issue can't be forgotten.
However, the themes of conflict aren't some that any one character can completely cover with overlapping at all. Unless the engine is much more abstracted, and there are only a few roles, and any of your checks ever are your "role checks".
Thus far, in my continued dissection of Sonshi's "Art of War" military forces seem* to break down into the following six roles:
-Assault/Defense
-Missile
-Command/Communication
-Reconnaissance
-Observation
-Supply
*:
"Seems" because he doesn't talk about pikers and archers (but does use the character for "crossbow" and "crossbow trigger" as part of his phrasing). However, he explicitly, and emphatically compresses "engaging/'conventional'" and "flanking/assaulting/'unconventional'" forces as having to be a "vice versa" unit type that have to be able to switch between both roles. He talks about archers. The break down of an army from duos & trios (specifically duos & trios, which must have meant something whose meaning might be lost in modernity); to squads, platoons and upwards to armies. To how commands and orders are to be communicated. The use of sentries and war camps. The employment of Espionage units. As well as the fact that he regarded armies as "self supplying"; where gathering wood and water, cooking food, and other support roles were filled by the soldiers themselves; not camp followers or support troops.
Most characters in a D&D game get a mix of them; and giving one character archetype arbitrary "role protection" is just that, arbitrary; and is going to cause more long term problems than not.
Someone can say that a "Warlock" who summons a powerful Demon should have their "Has a powerful Minion" schtick, protected; but that's a false argument. It also means that the character with a lumbering war-pet is suddenly "not allowed" (and I've seriously as many players with war-mounts in 2e and 3e games as I've seen NPC wizards with demon minions).
What you're arguing for isn't "role protection", but "theme protection"; the ice mage shouldn't expect an other character to pick up Ice Mage training, and somehow be 'better' than them. Which is fair; but isn't what I believe would happen with the method I'm doing the equivalent of throwing pasta at the wall.
The method that I'm imagining is more where the characters are picking either different Roles that they are picking up (and shuffling roles among the party; or trying to cover different bases) or they are changing their Themes for organic character development reasons.
Perhaps a better method is to have "type" and "power source" be separated along the lines of City of Heroes which would offer both theme and role protection.
Perhaps an other method would be for the basic roles that a character can take up have 20 levels of progression; and characters pick a power based on the dynamic they want to progress for their character concept.
And we make the aesthetic elements entirely window-dressing; or grouped and seperated so that the "broken" combos between a Role and a Theme can't be crossed (also, as in CoH).
So, suppose that there are six Roles
Force
Missile
Command
Intelligence
Observation
Supply
And that there are four Themes [borrowing from After Sundown's Universal, Infernal, Astral and Orphic]
Heroism
Pain
Madness
Death
So, assuming that at first you need Roles to be 10 levels deep; and each Theme gives something to each role.
You could have 10 * 6 * 4 = 240 options; and any character could be one of 12.96 million combinations of powers.
I'm not saying that this is possible, or should be done; just looking at how the issue of roles, themes, and character progression could be examined and evaluated.