Closing the bleed valves allows the engine to produce max power - which is generally handy when you've just lost half your available thrust.
Any theories as to the ~3 sec delay between the increase on #1 and autofeather on #2? A previous poster said 'normal', then corrected himself to say only when WOW... I also recall reading that somewhere myself. WOW appeared to be sensing correctly on the FDR.
#2, from the data on the trace, appeared to be running just fine until it decided to autofeather - thereafter it appeared to continue to run just fine, although at 'idle' and prop feathered.
There's more than one mystery here to solve I believe. Why #2 AFd in the first place, and why someone / something shut down #1.