Ycore Rixle said:
Do it up by the book. 1-100 creatures of random power levels appear right next to the caster.
Which book is that?
It is not the 1st ed. Advanced D&D
Players Handbook, in any case. That reads:
There is a 10% chance that 1 to 100 other creatures will be freed from imprisonment at the same time if the magic-user does not perfectly get the name and background of the creature to be freed.
There's no business about "intonation", much less the (potentially useful) effect you describe above.
The
Dungeon Masters Guide indicates how to generate the number freed (using a curve of d% x d100, so the actual chance of 100 is only 1 in 10,000 and the average about 26). It also states
For each such creature freed there is only a 10% chance that it will be in the area of the spell caster.
The level of monster is weighted by turning to Table IX (the penultimate) for 60% -- and level X not at all (so no solars, liches or yagnodaemons).
Even by an interpretation of the word "these" in the DMG commentary as indicating
all freed creatures (rather than only those in the caster's area), and strict adherence to rolls for type ... One could still fill slots a la your option (b) (just not with that being the sole determinant)!
The idea, though, that the random encounter tables should indicate the likelihood that
particular, known entities should (or should not) be among those freed does not quite satisfy me.
I mean, if one has established that (say) Leroy Brown (a human of class
x and level
y) is among the encysted -- in fact, that he is one of a certain assortment of creatures stuck in the locale -- then one might not be dealing with an absolute imponderable. Just how, in detail, the magic functions or malfunctions is not a subject the book treats.
I would not call it "railroading" if a DM, possessed of such established facts (which indeed seems likely come PCs of such level) came up with some basis more reasonable to her or his own sensibility. Enough
consistency for players to have the potential to inform their choices in general terms (an admittedly fuzzy standard) is desirable.
Still, I would agree that letting the probabilities play out is another step away from imposing a personal bias.