My comment regarding "good" megabeast was made with the understanding that we were discussing short term measures to tide us over the big wait and without new mechanics. In that context magic is out of the question, of course. While intelligent megabeasts probably aren't, quest giving probably is, so rampaging is about the only thing megabeasts are good for currently.
In the longer term, yes, magic powerful beings are definitely reasonable, as are intelligent quest giving powerful beings.
As for dragons, I'd expect we'll see worlds where dragons are primordial animals and others where they are powerful, intelligent schemers (Shadowrun style?), some of which may have taken on some kind of protector role. Yet other worlds may have both intelligent and beast dragons depending on sub species (where sub species and sub species properties probably would be generated procedurally). A problem with dragon species, though, is that it's hard enough for them to match up and procreate currently, so it would be very hard if split into species. Races (i.e. dissimilar subspecies being able to reproduce) might make it possible, as would mate finding behavior in larger worlds if the number of variants present in that world is kept down to a reasonable level (portals with off world feeding of the pool would be another possibility).
As for spheres, I'd use some kind tags to indicate which spheres sphere aligned critters are always aligned to, may be aligned to, and possibly sphere's they are opposed to/can't stand (an example without any particular application in mind: a creature that may be/is aligned with good and and may be aligned with water, but cannot handle cold. Such a critter/plant could be found in non freezing biome that are either good or non deserts, and may require both spheres to match if aligned with both). For each world where such creatures are found, world gen would lock down which one(s) of the possible spheres the critter aligns with in that world.