> 	(Presumably this is PR 3769).

That would be the one.

> 	Hmm - can someone say _why_ this can't be commited, or at least
> 	added as a kernel option?

It's a mystery to me.  Certainly leaving the timeout as is doesn't
render the code any cleaner than changing it to a more reasonable value.

In looking at the PR database just now, I see that PR 4765 describes an
ffs_valloc() panic during heavy fs activity to a ST34555.  Deja vu.
I'd be willing to bet $20 right now that it's the same problem.

-Steve