As you killed the broker it needs to rebuild all the indexes for the message store from scratch. In an ungraceful shutdown this is the behaviour you most certainly want.
There are always improvements being made to the persistence adapter so its worth trying the latest version if you are not doing so already. In addition you might try kahadb as this may have quicker recovery times.
I just noticed in the log your already using kahadb. One other thing that may help is to look at the hardware and see if a faster disk would give you more throughput.