Central questions: what does it mean for a spell to be balanced at a particular spell level, in the presence of shifts in play? What would a spell list / system of casting that really took the notion of shifts in play to heart look like? How would you have to change the resource model to balance it without breaking the world?
Contents: speculative proposals.
I've been kicking around a couple of projects lately. Both would entail designing new spells and figuring out what it means for a spell to be level-appropriate.
First, Magic the ACKSening. I would like to hit the old Arena-era feeling, with physical objects that a caster attunes to which let him cast a particular spell. I think there's some room for interesting mechanics here (blue card draw? Attune to a talisman in your possession temporarily. Black discard? Temporarily disrupt someone else's attunement). Also makes copying spells between casters within the party harder, makes competition for spells between casters in party more prominent. But, comes with a whole bunch of problems, like "how do you model the different colors? What's your spell-slots / mana pool mechanic look like? How does recharging work? How do you not shaft multi-color casters (if they even exist)?"
The ACKSiest way to do this would be to use the custom magic system creation rules from Axioms, but that makes life very difficult for multi-color casters. If red casting and white casting are two different magic systems, they track two different repertoires and sets of spell slots, and if you want to combine them in one class, that class is not going to be very good at either of them. In effect, its spellcasting will not actually be "level-appropriate" (I hate to use that word, but it serves), in the same way that a nightblade who casts as a wizard of half his level isn't actually going to be a useful spellcaster outside of the first couple of levels, or a high-level venturer isn't going to be a useful spellcaster. Neither has access to spells which are relevant in their current tier of play; they don't keep up with the subtle shifts. That's a damn shame, particularly in ACKS where those subtle shifts are not so subtle and a big selling point.
(This is probably a little too harsh on nightblade - if you pitch it as "at low levels when thief is weak, you're most of a mage with d6 HD and leather armor, and at high levels when thief is strong, you're most of a thief but with d6 HD, free Acrobatics, and the ability to cast Invisibility on yourself", it's probably a reasonable class. But my players sure don't use it)
I think this is fundamentally a problem with ACKS' design systems. They are too quantitative and insufficiently qualitative. For another example of this, consider fireball, which by ACKS' design rules should be a 5th-level spell. As I argue in the linked post, fireball is fine at 3rd - provided that the "expands to volume" clause is relentlessly enforced, in which case it becomes an outdoor spell, providing a big hit against either a humanoid warband (where the area of effect is the important point) or a single tough monster (where the d6 per level is the important bit, to force a morale roll through half-HP damage). It's much like sleep, which in the low levels provided the ability to either neutralize an encounter's worth of weak humanoids or to take out a single big monster like an ogre. The parallel is striking. I'm a little surprised that death spell doesn't have a single-target mode (which was split out into disintegrate), following that same pattern.
The second project I have been considering is a result of my travel to Taiwan for work last summer. While there, I visited the shrine of the Xia-Hai City God of Taipei. I was rather tickled by the notion of a City God. I subsequently read Journey to the West, and enjoyed its portrayal of the Celestial Bureaucracy, with deities ranked. So I'm considering attempting to model that sort of messily syncretic polytheism, with ancestor worship, immortal heroes, animal spirits, boddhisatvas, dragon kings, hearth gods, stable gods, all manner of gods being worshiped, preferably with a minimum of systematic complexity. I'm considering keeping just one divine caster class. Your first and maybe second level spells are from ancestors and commoner deities like the Kitchen God, your third and fourth from regional deities, and then your fifth level spells are from higher powers, picking (say) two of three patrons per spell level, which then determine your spells available at that level. But in any case, would need (many) new spells for this as well, hence the interest in "balance" with existing divine options.
So there's the question - if I'm dubious of ACKS' numbers for producing really balanced, tier-appropriate spells, I need to get a better qualitative idea of what's appropriate for a spell of each type (arcane | divine) at each level. I think I was on the right track here - generally arcane does just straight-up solve (or oversolve) a level-appropriate problem at low levels, while divine does tend to only half-solve problems, give a bonus, or something like that. But there's more to it than that. I'm interested in the relationship between eg Wizard Lock and Hold Portal, 2nd and 1st level arcane spells that do roughly the same thing at different degrees of thoroughness. That might seem like a silly thing to care about, but I'm kicking around spells that fog whole hexes to solve fundamentally the same problem ("we are being pursued") at wilderness levels, because the wilderness is just a big megadungeon - so if your wizard can have a spell to solve the problem of pursuit in the dungeon, and now we're doing wilderness stuff, well sure, you should have spells to solve similar problems at this new level of play. Flood Ford and Hold Mountain Pass as 3rd-level analogs of Hold Portal, maybe... Light solves the problem of tracking consumable torches temporarily, Continual Light solves it more permanently - Create Food and Water solves consumable wilderness resources temporarily, Cornucopia solves it more permanently by imbuing an object and requires maintenance (the divine version, meanwhile, is more like Call Game Animals and gives you a bonus to hunting throws to find food. Purify Food and Water / Preserve Food and Water are also more along the divine "partial-solve" lines, but rely on enforcing ration spoilage).
If wizards could straight-solve wilderness-level problems like they straight-solve dungeoneering problems, maybe wilderness play would be more fun. Perhaps my players would be willing to take more risks if they felt they had some aces in the hole besides fireball.
Unfortunately, this [high-powered wilderness- and domain-grade spells] is in tension with ACKS' ethos of "let's tone wizards down a little so that we can still have conventional armies and historical-looking economics." This is why ACKS' fireball only has a 10' radius instead of 20', for example. A partial resolution to this is really enforcing the "spell slots only recovered by rest in safe, sanity conditions" rule - this means that yeah, that wizard is gonna wreck one unit, but then he has to get back to town to recharge. This breaks down under siege conditions, providing a huge advantage to casters defending a town over those laying siege to it, and might have other unforeseen edge problems. Maybe the right thing is really to vary the recharge time for spell slots by phase of play. Dungeoneering-tier spells recharge every day, while a turn in the dungeon is ten minutes - a similar scheme would have wilderness-tier spells recharging maybe every month, against a "wilderness turn" of one day (... or maybe every 3 months; 144 10-minute turns per day, and 144 days is 4.8 months, but that seems excessive), and domain spells... I don't even know. The upshot of boosting the recharge times is that you can go nuts with high-power spells that solve level-relevant problems but aren't really useful at lower tiers (does raise some questions like "how do they interact with Dispel Magic?").
Maybe it makes sense to move the bar for ritual spells down. The nice thing about spell slots, though, is that they don't build up over time like ritual spell items do - lord help him who goes to war with the elven or lich wizard-lord who has had a couple of centuries to stockpile ritual items. Maybe a limit on "sustainable" ritual spell items, just like on Continual Light? Or you have a spell slot, but you have to do something like ritual research to recharge it? The logical endpoint of this train of thought probably looks rather vancian... I think I like longer recharge times better.
There is also the difficulty that most of the existing 5th-6th level spells are really more like high-level-dungeoneering spells than wilderness or domain spells. Passwall, disintegrate, anti-magic shell, feeblemind, dimension door... this stuff has domain applications (siege, fighting high-level NPCs and big monsters), but it's not for domains. Passwall is for punching temporary shortcuts in dungeons; disintegrate's secondary function clearing a 10' cube is the same but more permanent, and probably wasn't designed or balanced with reducing fortresses in mind. They're not designed around the shifts, and if your game doesn't include dungeons as a salient feature of high-level play, they're a bit out of place (especially if competing for repertoire space with eg a spell you can only cast every five years that summons 500 demons for a single mass combat).
Maybe tacking these high-level dungeoneering effects onto low-level spells when cast by high-level casters is a reasonable solution? So say, knock can produce passwall if the caster is 9th+ level, much like bless can make holy water or magic missile scales up. And then 1st-2nd level spells are your dungoneering and personal combat bread and butter, 3rd-4th are mostly wilderness or small mass-combat (haste, prayer, and other mass- spells fit here) with longer recharge times, and 5th-6th are for domains and big mass combats and have looong recharge times.
What a mess.
This is starting to sound suspiciously like Trailblazer's reforms to spells, grouping them into different recharge times even within the same level... And I also adopted their solution to half-casters being bad. Trailblazer was right about everything. Well OK, maybe not action points, but a lot of things. Although even ACKS is headed in that direction with the Heroic Fantasy book...