I think if you've got the cash, do option 2 - (come to thing of it prob cost more for enterpise edition ) youll get fantastic results from having the logs on a seperate controller -especially if there is nothing but the logs - then it can just write sequentially - disk spindle never loses its place.
Maybe also just get a RAID 1 array - wont have the overhead of striping.
on the worker threads - havent done it myself, but i reckon it pretty much boils down to your server resources - maybe up them to 500 and see. (or better - see if you can do a trace/pf monitor on them & see if they are maxing.)
Have heard the same about boosting - never dared to do it! though our servers never run only sql - so it would be crazy for us. is yours only running sql? prob not too good an idea to do it on production server
Have you set Use Windows NT Fibers? Helps with the OS & system management...
also - if you end up getting a nice new pile of disks - throw your heavily used table indexes on separate spindles to the data files too- will help a lot. (as a compromise, maybe throw them on the new log array)