"But doesn't having a notion of 'balanced' combat inherently imply that all combat encounters are expected to be fair and winnable" well, no – it implies only that the GM has the ability to know whether a given combat encounter is fair and winnable.
There's a story that's been going around for decades about a Dungeons & Dragons party who encountered a large room full of treasure while exploring a dungeon. Immediately suspicious, they asked their GM a series of detailed questions about the room, but no obvious dangers were identified. Satisfied, they moved into the room – and were immediately set upon and eaten by the dragon that had been sitting atop the pile of treasure the whole time, which the GM hadn't mentioned because the players never specifically asked about the presence of living creatures within the room.
While this is obviously an extreme and ridiculous case, it illustrates an important point: as GM, you're the group's eyes and ears. If you don't describe something, the player characters literally can't see it – that dragon was effectively invisible from their perspective. The trick is that active malice isn't the only way to invisible-dragon your players; a group can also find themselves invisible-dragoned because the GM simply failed to provide sufficient information for the risk in question to be identified. This can happen through neglect, but it can also happen because the GM themself was unaware that the risk was present.
Now, hold on, you might be saying: the GM "plays" the entire world. How is it possible for the GM not to know that a risk is present? Well, that brings us back around to the subject of combat balance.
A game in which "balanced" combat is a meaningful thing to discuss is typically going to be one in which both the players and the GM are actually making strategic, tactical, and/or logistical decisions, rather than merely producing a description of their characters making such decisions. Without a good handle on the interplay of these decisions, it's completely possible for the GM to be wrong about the level of risk the scenario they've constructed entails.
That's actually pretty critical, because even if you don't care about the game being fair and winnable (and that's a perfectly valid stance), your players are still depending on you to be their eyes and ears, and to give them enough information to make good decisions about whether the fight in front of them is one they can win. A game where not every fight is expected to be winnable needs to be a game where the players have the opportunity to walk away.
No matter how objective you try to be, your own sense of the answer to that question is inevitably going to colour how you communicate about it. You being wrong about the level of risk at hand inherently increases the chance that your players will make bad choices. The party eating a TPK because they made a stupid decision is one thing; the party eating a TPK because they made a decision that looked reasonable from their perspective based on your unwitting miscommunication of the level of risk involved is quite another!
Sure, once the dice hit the table I'm probably going to realise that I fucked up, and I can adjust things on the fly to bring the level of risk that's actually present in line with the level of risk I communicated – but that's extra work I don't need with everything else that's on my plate. And that's a best-case scenario; if I'm running the game for a hardcore let-the-dice-fall-where-they-may group (and such groups tend to have a pretty significant overlap with groups that are cool with not every fight being winnable), I may not be able to adjust the fight's parameters on the fly without violating the social contract of the table.
Basically, whenever I see an OSR game with tactically crunchy combat brag about how its author never even thinks about "balance", what that's telling me is that running this game is going to create a whole lot of extra work for me as a GM. This is not a selling point.