In this setup, individual spells are rare, restricted to only a very few spellbooks in the world. But the range of spells is wide. While as a GM, I'll be crossing off spells when I roll them up to place in spellbooks (because the spellbooks found in the game will have random contents), every single spell in the game is possible. I'll just check off entries as they come up. This means that my NPC wizards will have strange and unpredictable spellbooks, so I'll have to be creative in how they use those spells. It also means that as the campaign progresses, PC spellbooks will grow and change in semi-random ways, though the players will at least be choosing which spells to try to learn from the ones they find. The breadth comes from the game books -- every spell in the 4 volume Wizard's Spell Compendium set is available, and I'll be rolling them up from there.
In the comments of my earlier "one of a kind" post, Kent pointed out that powerful wizards could reasonably be expected to have gone out and collected all the spells. This goes hand in hand with the "copying problem" -- unlike magic items, which can be passed from hand to hand without reproduction (perhaps simply by stating that the secrets to their production have been mostly lost), spells are different. You can easily copy of a spell from a spellbook and add it to your own.
One check against this kind of thing happening is that I'll be using a modified version of the "Spells Known" table from Holmes Basic. The original table is here. The table is great, but it's designed for a version of the game with only about a dozen spells per level. Using the Wizard's Spell Compendium set means that I'll have hundreds of spells per level, so that jump from a maximum of 10 spells known at Int 16 to a maximum of "All" at Int 17 becomes a hell of a jump. I'll probably be changing the max spells for Intelligence 17 to a maximum of 12 known, and Intelligence 18 to a maximum of 14 spells known at each level.
This really does force the wizard to choose. What spells do you study and attempt to master, so adding them to your spell book? Which do you not? When PCs find a new spellbook, the wizards will have some hard choices to make. This does raise the question of what to do with the other spells that aren't learned and added to the spellbook. The answer, it seems to me, is simple: you tear the pages out, and you have an instant scroll. And once you cast the spell, it's gone for good, consumed by the casting.
This means that wizards will generally be able to master only a dozen or fewer spells. They'll have to pick and choose when a spell is worth putting the effort into learning. What's more, a lot of spells were lost when the world ended. To collect these spells, wizards need to first find out that they exist (through rumors, research, and so on), then find out where they were last known, then track them down and maybe find out where they now sit in a lonely ruin with the bones of their former master. And then you still have the task of going out to get them. Also, because it's a low level world -- there won't be many high level characters alive out there, there's a reduced chance that any cabal of master wizards will have collected all the good spells.
Kent also suggested that maybe the act of copying spells might itself be intrinsically dangerous, and I like that. A "spell copying misfire" table could be fun, or I could go with one of the "wild magic" tables from the 2e books, possibly limiting the effect to times when the "chance to learn a spell" roll is a failure. I'm tempted to also use this "chance to learn" percentage as a casting roll for casting unknown spells directly from scrolls and books. Maybe a failure on this casting roll would call for a misfire roll as well. I don't want to make magic so dangerous and unpredictable as to be more trouble than its worth -- but I do like the idea of keeping it strange and wonderful.