Hi liangstein,
Yes I've started seeing that too since moving to the new SDK.
I've changed it to exit on that error for now so you can restart and keep going.
When I've seen that error I haven't got the part about e_alloc - that's really weird as e_alloc should only ever be called once!
I'll have a look later, but for now I suggest running it in a loop, or using the instructions I've just added to the readme to run the miner as a service that reboots the parallella after any error (obviously only do that if you aren't using the parallella for anything else!)
Cheers,
Mike