I dunno. I could say the same thing about 5e rapiers versus other one-handed weapons, or even archery versus other fighting styles. If not broken, they exist outside the normal power curve, and I would certainly say choosing those options counts as optimization.
My view here is that if there are broken elements within the game (or ones that exist outside the normal power curve), then choosing them is reasonable: they're on the menu. If choosing them will break the game (or cause overshadowing issues, etc) then the answer is to change them. Which is easily done if they're discrete elements. For instance, the katana can just be changed to a 1d8 weapon.
In Prince Valiant starting as a knight is an option. So is starting as a squire or man-at-arms. A knight starts with 800 Fame, the latter options with 500. The former is therefore closer to getting the first skill improvement (which comes at 1000 Fame). Here is what p 57 of the rulebook says about this contrast:
The advantage of playing a squire or man-at-arms is the lack of entanglements. Knights must be concerned with loyalty to their liege lord, leadership of underlings, good manners, the code of chivalry, and other kinds of responsibilities, and those knights that ignore society’s rules are reviled. Squires are less harshly judged. If your Adventurer is a knight he is often the center of attention, while lesser-ranking warriors are not so carefully scrutinized, and can act less formally. So if you would like to have an Adventurer who is a warrior but is less concerned with proper behavior, take the occupation of squire or, better yet, man-at-arms. . . .
If your Adventurer starts with the occupation of squire or man-at-arms, he can be promoted by any knight to that rank in an exciting, dramatic ceremony. Of course, some great deed of leadership or courage must be achieved for knighthood to be conferred.
In our game, two PCs started as knights; a third player's first PC was a knight, but that character died in the first session (dragged down to Hell by the Wild Hunt) and he started his second PC as a man-at arms (who has since been knighted). Choosing to be a knight can hardly count as
optimisation, can it, in any meaningful sense? There are two choices, the difference between them is crystal clear, and one chooses one's preference. Being 300 Fame behind is about one session behind in a system where PCs start with 9 skill ranks and get a new rank every 1000 Fame.
If a group doesn't want this difference to be part of the game, they can just insist that everyone (or no one) starts as a knight.
Burning Wheel has much more intricate PC building than Prince Valiant, but in some respects it is similar: there are ways of building characters who start with more gear, more reputation or more skills than others. It's assumed that players can see this, and - a bit like the passage I quoted from Prince Valiant - will have a reason for choosing to build a different sort of PC, if that's the choice they make.
This also closely relates to the issue of win conditions . . .
I do think asking what play priority is being violated by optimized characters in a standard adventure path type game is an interesting one, though. Everyone playing is aware that they're going to eventually win out (even if their character dies, you make a new one), so what does it really matter if the combat is easier and goes faster? I feel like it's probably the performative aspect being negated in combat by the stronger characters, but I'm still thinking about it.
I think this bit of your post that I've quoted rests on an assumption that
the successful play of the game involves moving through a series of combat encounters and that
optimisation is about better controlling the outcomes of those combats and winning them more handily. As you flag, this has knock-on implications for pacing, for what other activity occurs at the table during the play of the game, etc.
If combat encounters are
just another thing that might happen in play and are not connected in any particular fashion to progressing a PC's (or player's) goals, then the idea that combat performance is something that there might be special reason to enhance, or special reason to care about, evaporates.
In our Classic Traveller game, there are a large number of PCs, some "primary" and some "secondary" (comparable to retainers/henchmen in classic D&D). Of the four
most primary PCs, one can shoot a laser carbine but has done that, I think, two or three times in 20 sessions of play. The rest of the time his contributions flow from his knowledge of xeno-archaeology and his expertise as a medic, administrator and starship navigator. A second is a playboy baron whose main contribution to the party is money and a starship. A third is a spy who can shoot a shotgun if he gets desperate but I think has never done that in play. And the fourth is really handy with a cutlass, and uses that maybe one in every three or four sessions. She is also a (currently rather weak) psion - if/when she maxes out she'll be able to do a bit of self-buffing and clairvoyance.
There's no real correlation in this game between
areas of performance that one might try and enhance and
win-conditions or anything in that neighbourhood. Of course anyone can look at the secondary PC with 17 skill ranks and stats DA3CD6 and see that she is (in that sense) stronger than the playboy with 5 skill ranks and stats 4A7A6C. But while these two characters are in the same player's position, it is the playboy and not the secondary that he chooses to focus on. Because that is the character whose "story" he is invested in, who he has goals for, and around whom I'm (therefore) framing fiction.
This is the same group I played 4e D&D with for six years. But the different system, which includes a very different mechanics-to-fiction dynamic, means that optimisation isn't really a thing. (Whereas I have no doubt that it is a thing in 4e D&D - though even there you can choose to optimise for combat, or for non-combat.)