Actually what if we took that side effect idea and made it into a retort?
Spellweaver 2 picks
Get 4 fewer common and uncommons spells, but 4 more rares and very rares.
Basically, for 2 picks you could get 2 books which are 2 rares, 2 very rares, 2 uncommons and 2 commons. (not always but on average), or with this, 2 more rares and very rares than from books, but you lose your early spells.
...this doesn't sound very fun either.
What about "You get 2 more very rare spell for each of your realms that have 2 or more books in it.", specifically intended for 2 or 3 realm wizards? We do have omniscient for early game multicolor, but it doesn't really work for late game (you had to give up a pick on the retort itself, and extra realms to make it useful, assuming you did go for the multi-realm effect and not just for one or two of them).
So with this retort you could get 4 very rares for 2 picks, which no books can possibly give you if you have 2 realms (but 1 realm gets it, as the last 2 books have exactly that amount), or, could even get 6 very rares, if you do 3/3/3 or 4/3/3 books, but that's the most. Maybe to make it even more attractive, throw in a smaller, 6-12% cost reduction for rares and very rares, or an extra rare spell.
(However, this would be a retort that should never appear in treasure (or if it does, should add the extra spells), and the AI should never pick it unless playing 2 realms. )
So summary, it would be
Spellweaver, 2 picks
Your rare spells cost 6% less to cast, your very rare spells cost 12% less to cast. You can research (an additional rare and) two additional very rare spells in every realm you have at least 3 books of.
...so basically it would be as good as the 9th and 10th book pick, without it having to be those books, but the catch is, the effect is limited to the late game. Might be still too weak? We could do 10 and 20% instead maybe? Or is that now on the too powerful side?
Spellweaver 2 picks
Get 4 fewer common and uncommons spells, but 4 more rares and very rares.
Basically, for 2 picks you could get 2 books which are 2 rares, 2 very rares, 2 uncommons and 2 commons. (not always but on average), or with this, 2 more rares and very rares than from books, but you lose your early spells.
...this doesn't sound very fun either.
What about "You get 2 more very rare spell for each of your realms that have 2 or more books in it.", specifically intended for 2 or 3 realm wizards? We do have omniscient for early game multicolor, but it doesn't really work for late game (you had to give up a pick on the retort itself, and extra realms to make it useful, assuming you did go for the multi-realm effect and not just for one or two of them).
So with this retort you could get 4 very rares for 2 picks, which no books can possibly give you if you have 2 realms (but 1 realm gets it, as the last 2 books have exactly that amount), or, could even get 6 very rares, if you do 3/3/3 or 4/3/3 books, but that's the most. Maybe to make it even more attractive, throw in a smaller, 6-12% cost reduction for rares and very rares, or an extra rare spell.
(However, this would be a retort that should never appear in treasure (or if it does, should add the extra spells), and the AI should never pick it unless playing 2 realms. )
So summary, it would be
Spellweaver, 2 picks
Your rare spells cost 6% less to cast, your very rare spells cost 12% less to cast. You can research (an additional rare and) two additional very rare spells in every realm you have at least 3 books of.
...so basically it would be as good as the 9th and 10th book pick, without it having to be those books, but the catch is, the effect is limited to the late game. Might be still too weak? We could do 10 and 20% instead maybe? Or is that now on the too powerful side?