The most important practical difference is that Power PC is not multi-copy atomic. That is to say, "[the PPC] memory system does not guarantee that a write becomes visible to all other hardware threads at the same time point." AArch64 does guarantee this property, although it was not documented as doing so in the first release of its specification. For this reason PPC has to emit some extra StoreLoad memory fences.
(See "A Tutorial Introduction to the ARM and POWER Relaxed Memory Models" by Luc Maranget, Susmit Sarkar and Peter Sewell, but be aware that the AArch64 specification has changed since that paper was written. Our work has resulted in Arm clarifying the AArch64 specification in a couple of places, for which I'm grateful.)
It worked far better than I expected. Around the turn of this century I was involved in bringing up the software for a new processor made by a certain large semiconductor company (you might be able to guess!) The early hardware didn't work perfectly, the operating system wasn't ready, and the compilers were flaky. The cost of writing a simulator was cheap when compared with avoiding all of that pain and getting first-mover advantage.
No, I'm pretty sure there's no intention to restrict redistribution. The "services" refers to support, I think. I'll see what I can do to get this clarified.