Will it be merged?

There is opposition to merging realtime preemption
The patch is intrusive
Code becomes a little more obscure
Even 2.6 kernel preemption not entirely trusted
Big change for a small user group

Parts have already been merged
Lock breaking, spinlock initializers, completions...

The rest may drift in over time
Each part will have to justify itself
Value to most kernel users
No cost where not used

Expect some big fights
And possibly even a 2.7 kernel