A few answers you requested:
#1- Blend is just fine.
#2- Blend test is again fine. However, if you want to get bonus points you can run MemTest for a few hours to make sure your RAM is stable.
#3- Running 1:1 is preferred. Running odd multipliers might create a few issues, and you will not notice a performance difference. Technically speaking, an increase will be there, but it will be very negligible and you'll only notice a difference in synthetic scores. Real world applications and usage won't be noticeable.
#4- FSB hole is well.......the point where your OC takes a dump into the black hole of despair and cannot find its way out. Generally speaking though, if one does exist, you SHOULD be able to just jump to a higher FSB and get "out of the hole" and continue OC'ing. For example, a certain mobo might have a hole at 375-395 FSB. So chances are if you just jump to 405 FSB then your OC will actually be stable (assuming you have the voltages working correctly and stable) but it won't be stable between 375-395. Go figure. It's not terribly common though, but I have seen it happen on occasion.
#5- Prime95 will stress your temps far more than any "normal" usage and applications would. It's designed to push your hardware to the limit to ensure 100% stability. Normal daily use will not push it that hard, and certainly not for 24 hours. Maybe a few minutes at most, unless you're doing intensive tasks regularly for long periods of time such as graphic renderings, animations, etc.
#6- Use several tests to measure the differences in performance. No single test will really give an accurate picture overall. Everest, Sandra, and 3DMark Vantage are popular synthetics. Beyond that, for real world, you can run Super-Pi, WinRAR, Office, etc with a stopwatch and chart the differences if you like.
Hope that helps.