Roywyn

Test from a guildy with untalented Arcane Missiles (no crit modifiers):
598-599 hits. 924-925 crits
924/599 = 1.54257, 925/598 = 1.54682

Looks exactly like 150%*1.03, a flat 3% increase of the final crit damage.
We had the same result with fire (Molten Armour 89 hits, 138 crits, but the damage is too low for exact predictions).

So, we have a normal effect on spells without crit damage modifiers, but an increased effect on spells with crit talents.

Looks like the resilience issue to me, that the crit talents apply to a pretty odd number for the crit bonus damage:

C := {(150%*1.03) - 100%} is the difference between a normal hit (100%) and a crit with the crit gem and no crit modifier talents (150%*1.03).
(( Resilience works the same way, with the resilience modifier (0.9 for 10% crit damage reduction) instead of the listed number 1.03 (crit gem). ))

The final damage for a crit seems to be (it fits all test results and fits the resilience problem)
100% + C*(100% + crit_talents).


So, for spells without crit talents we get 154.5% crits, for spells with a +100% crit talent (elemental shaman, balance druids, destruction warlocks, frost mages) we get 209% crits.
Arcane mages get 181.75% crits, arcane/frost mages get 236.25% crits, fire mages (Ignite is no crit modifier but just multiplies the total damage done by 1.4) get 216.3%, arcane/fire mages get 254.45% crits including ignites.

My final thoughts - it's much better than the spirit shard gem, not as good as the half cast time gem. The latter one has a stupid gemming requirement and can mess with spell timing (I have it on my T4 hat and it's really really fun!), so I'll probably have +12 agility on my next hat :o