I think it's pretty well established in any edition that the publisher can sell a pretty big number of Class books. Sword and Fist, Complete Warrior, Complete Champion, Martial Power... After the Core Rules, these are usually the next best sellers.
A bit less true in 4e. Actually, a lot less true in 4e - the DDI basically ripped the heart out of splatbook sales. However, since the DDI itself can pick up the slack, that doesn't negate the rest of your points.
Given that each class is likely to have that sort of treatment in a supplement, I think it makes sense to leave the Core Rules version very bare-bones. Playable, of course, but only just. Proceed with the assumption that players who want a deeper experience with a given class will invest in its splatbook anyways.
This may be one of the big issues WotC face with 'reunification', but this is the absolute opposite of what I want. Pre-4e, I found that the splatbooks most often made the game considerably
worse, as material outside the core wasn't playtested enough and so was too strong, too weak or (most commonly) conditionally broken depending on whether the Character Optimisation boards had found a hack.
4e didn't have this problem to the same extent, because of the ease of errata/revisions. Unfortunately, those same errata/revisions were a
major stumbling block for me with 4e. Also, the game very rapidly suffered from option bloat.
My strong preference is to play "Core Rules Only". I don't expect that to change. So, I want those core rules to provide a smallish set of well-developed options - I only need 4-8 classes, but I want several options within those classes.
The alternative would be to produce multiple themed Player's Handbooks. Fewer but beefier classes in each PH. Which they tried, in 4E. I wouldn't be able to say whether that particular strategy was successful or not.
I like the Pathfinder model: a reasonable set of fully-detailed classes in the Core, then another 6 fully detailed classes in the APG, and then the "Ultimate..." books each with a few more fully-detailed classes. (3.5e had much the same idea, but a different order: PHB, then "Complete...", then PHB2.)
I think it is important to note that any class that isn't in the Core simply will not get the same support as the core classes. This was particularly note-worthy in 4e with Wizards, which seemed to get a huge amount of support. Therefore, if you are going to introduce classes outwith the core, it's probably important to fully-detail them at the outset, so that people aren't scabbling around for scraps of support later. (It would be nice to think that new classes would get proper ongoing support... but the reality is that they just won't. That's just economics at work - almost everyone can use support for the Core, but a relatively small subset can use support for their own favoured supplements.)