no it would run for usually up to 20 hours in one run at a time, and i had usually about five but sometimes up to eight instances of the process running at the one time (even with all @ Below Normal priority it was very noisy as it made all my fans go to full and computer is in same room as bed!), but yes had to run for over 3 months all up (approx 16/7 overall). this was image processing with varying size kernels 4x4 to 8x8, make 1 subtle modification to the kernel, reapply to testbed of 1000 images, do statistical analysis, repeat... 3 months later all kernels are analysed and the cream rises to the top
too scared to think how long it wouldve taken on my old pc which is like 8 years older lol, i wouldnt have even considered it. I wouldnt have even started the whole thing had i known it was going to take so long but after a month it was like ok dont want to just waste a whole months work, so kept going. It was a lot of work for little reward, but not a waste.
i was thinking ones with more instructions than can be prefetched (yours had about 30 bytes worth), ones that run a long time like the **** i was running Dec-Feb lol (yours ran for 1.4 seconds!), ones with non-predictable branches (yours had none), ones with nested loops (yours was one unnested loop), ones that read and write data, etc. No point wasting your time on this though, I only suggested it because i thought it seemed (im obviously just speculating) very easy for Fred to be able to add support for something which Intel recommends, as it's essentially just adding support for an "! align" statement to be inserted and it would complement PB's data/structure alignment, no biggie! 
