Scientific computing, system modeling, numerics, CAD/CAM, visualization, developing highly concurrent software, event processors, rendering, machine learning (esp vision/video/spatial stuff), genetic algorithms, databases, GIS, crypto, traffic processing, theorem proving, compiling stuff, exploratory data mining. Gaming, maybe. I routinely saturate my MBP's 4 HT cores. It's also nice to develop on the same architecture you'll deploy to; helps with concurrency tuning.

Not that you can't run these on smaller devices; these are just easy ways to use up a lot of CPU.

Good reasons but I wonder is the current/new generation 12-core Mac Pro inadequate for this? It could be better but it still seems to serve its purpose.

Depends on what you're doing. I'd be satisfied with a 12-core Mac Pro for everyday development--but if a 128-core Xeon were available I'd take it gladly.

When I was working on quantum state diffusion, it took many hours on a 24-node cluster for a single run. In many of these tasks, the problem will expand to consume all reasonably available resources; more cores allow greater precision, wider sampling of parameters, higher fidelity, etc.

some of the iOS developers at my work use them just so they can compile faster.

then they're doing it wrong.

what? Clang parallelizes compilation 6 > 4. duh.