SolipsismX, I'm a software engineer, most likely with more experience in operating system development than you, so drop the condescending tool least you find yourself humiliated. I know what a memory leak is.
The terminology isn't complex stuff. The word leak should be understood by all English speakers.
Good, because you make it sound like otherwise.
Is the memory never released back to the system?
Ambiguous yes / no question; both answers would be "no", with the correct long version being "no, the memory is never released to the system". Notice that I did mention that Safari does that even with all of its windows closed. And before you even consider it, no, it's not an add-on; disabling add-ons actually makes it worse.
Does it require a restart or to manually killing the process for the memory to be released?
It needs a restart, yes, but doesn't crash (just causes a lot of swapping when its memory usage grows to 12-16GB).
Does the system run out of RAM and never reduce the amount used by Safari in order to give more RAM to new apps?
What part of swapping did you not understand?
Have you tried to open up another memory heavy app like Aperture when you have less than 2GB free to see if Safari gives RAM back to the system?
Again, what part of swapping did you not understand? Safari eats RAM that does and does not exist; it will happily eat my entire hard-drive's worth of swap if I let it.
What I've seen with people with RMBPs is that Safari is using a lot of RAM but they have moved from machines with an average of 4GB RAM to machines with 8 or even 16GB RAM. Mac OS X will allow a lot of apps, especially native OS apps, to hold a lot of RAM to make it perform better but if another app needs it more then that RAM from that process.
You are confusing a lot of concepts. For starters, what you are most likely talking about is the paradigm in which applications are never closed but rather moved to the swap when they are unused and if their RAM is necessary (all operating systems do this if you don't close applications). This, however, does not give applications a free pass to privately allocate as much RAM as they wish, because those allocations are still wasting resources, and the amount of swapping that they cause slows everything else down as it causes the system to move other stuff to the swap sooner. There are also other cases in which all operating systems manage RAM intelligently as long as it's available, such as disk caches and shared memory maps, but we aren't talking about either of these here.
You have shown how much RAM WebProcess and Safari were currently using but you've shown no evidence of it leaking, a action that shows a change and has some detail as to what was going on between the two data points. I see WebProcess is using 8.4% of CPU. That's quite a bit which means it's actively running not sitting idle and holding onto RAM that the OS can't delegate to other apps.
Of course it's running; sleeping processes can't allocate RAM...
That is not to say you and others do not have a memory leak but you nor anyone else have shown any tests that would prove your hypothesis. Until you can do so you shouldn't be spouting your assumptions as fact.
I don't think you understand the subject you are debating.