Yes, the contrast between key-on velocity on one channel and crossfade on another gives you more possible expressions.  I have found that the crossfaded timpani cresc/dim is absolutely indistinguishable from the recorded cresc/dims in any orchestral useage.  So you have a lot of variety possible there.

Concerning the strings, I am wondering if anyone is doing divisi with contrasts in noticeable size between the various string ensembles.  I have found that creating a matrix with Appassionata 20 players (in violins) that splits into two part Orchestral violins (14 players) sounds almost perfect as a 2-part divisi.   Even though the numbers do not match precisely, the sound is exactly like a divisi of that original massive ensemble sound.  You have to reduce the volume of the 14 player slot to approximately 80 or the doubling will increase the level incorrectly.  Likewise, if you need three part divisi, you can split the Appassionata violins into three chamber violins patches (or the other strings analogously) and by correcting the volume on the smaller ensemble (which has three notes so is close to tripled in volume)  it sounds excellent.   You have the option of adding a solo also on any of the parts, since the numbers are still close (and therefore indistinguishable audibly) and it adds another layer of complexity to the mix.

Yeah I agree with William's divisi and layering ideas.  I did try once, layering solo violin on top of orchestral strings.  It was magnificent... but I didn't experiment too deeply. It added that extra layer of "a few musicians that stick out a little bit on an expressive passage" which is very satisfying to hear on lyrical passages.  The catch however, is that the solo instrument has to be mixed just right - its quite (pun blatantly intended) fiddly to get it to "blend" but "stick out" at the same time.

I wanted to point out a few other things, too, about VSL's dry approach versus other libraries embedded-reverb approach.  There are the usual arguments, that it allows us to place instruments anywhere in the stereo field, and allows us to place the instruments in different reverb spaces... but I think there are a few more profound points about dry recording that make it a brilliant design decision:

(1) VSL's legato is not just superior because it existed when other libraries didn't have it.   It is superior largely because they recorded *dry* legato.  Even though other libraries have finally caught up to the idea that they need to record performance intervals, they're still missing that dry recording is key.  Legato samples recorded with reverb still have some significant problems.  In particular, how do you stitch together the samples and deal with reverb tails of previous notes?  The only logical place to create that stitch, especially to support the composer's arbitrary sequence of notes, is just before the legato transition to the new note.  But when there's reverb involved, then the next note needs to somehow magically know what the reverb tail of the previous note should sound like.  Its extremely difficult to edit the samples so that legato AND reverb are simultaneously seamless for all possible intervals.

(2) Another fundamental detail that makes VSL vastly more expressive and versatile is that EVERY articulation can be shortened... not just sustain types.  Consider a staccato note (or worse, a detache note) with reverb embedded in the sample.  What if you want that note to be even shorter?  Then the software needs to truncate the sample (including that chunk of reverb), and then salvage the reverb tail they recorded... that stitch between the truncated sample and its reverb tail becomes difficult because the jump in timbre at the seam just doesnt quite match.  On the other hand, VSL can play shorter staccato notes easily, because non-reverb truncated sample and non-reverb release can be stitched together more easily.  But this advantage goes even deeper.  Adjusting the length of short notes becomes an additional way to vary the expression, in addition to round robin samples and velocity layers.  A well placed full staccato sample amidst secco, staccatissimo notes might add just the right amount of emphasis, whereas a new velocity layer doesn't sounding right.  

I personally felt that this aspect of expression wasn't there in other libraries, but its been (pun blatantly intended) instrumental in my mixes =)

I had not thought about the reverb embedded in legato transitions.  That is an interesting point and I think it may be true, since you actually do hear it as soon as the pitch rises or lowers enough.  Creating an additional reverb on top of reverb already there would then be muddier than with the clear, dry recordings which VSL uses.  I agree also that the dryness allows much greater freedom to alter individual notes. 

One thing like what you mentioned on the string layering that I've tried to do is when the sound of a section is loud, you hear almost entirely the ensemble.  But when it gets quieter, you hear a little more of the solo layered.  I've noticed that effect, which is obviously caused by the dynamic range of a single player vs. the entire group, on live recordings when you can hear more individuality on quiet passages.