Thanks for the reply, I appreciate it.
The slowness was when trying to apply the profile to a blade, with that application taking up to 35 minutes. I tried failing over to the second Flex10 and then back again. Each failover took a long time and didn't solve the problem.
Using the OA CLI to reset every server solved it. Profile applications now happen quickly. The symptoms and resolution appear to be similar to a supposedly fixed bug where nvram can be corrupted during certain power-cycle events. I guess it might not be totally fixed.