@agorrod, here are the changes to skip the in-order key check. I also exploded the bulk-load cursor routines into one function per type of insert, so this should be about as fast as our current bulk load implementation is going to get.

I'm going to run a quick performance test just to satisfy my curiosity, but you should test it on the workload you think is the right one – if it's faster for you and you agree with the approach, I think it's ready to merge.

Reference WT-1127.