I believe that suppressing components has the same effect as deleting them. I've never actually tested that out. Since you have a lot of suppressed components in the above assembly, you are in a very good position to test that question, by using Pack and Go to copy the above assembly to another directory and then deleting the suppressed components.
I'm afraid I can't help much with your other questions.
I liked your idea to test this, I packed the assembly (which had changed a bit) and did the test both on and off my network.
Here are the results:
Regular - 1m 6s
Simplified - 28 sec
Regular - 13.5 s
Simplified - 3s
These are the times from opening the part to seeing the design tree (and being able to use it)
I was suprised at how much the suppressed items actually affected it, and also how bad using the network is.
I did not do any tests involving saving, though I know save as is significantly better than just saving.
Regular assembly properties:
Simplified assembly properties:
I almost forgot to do an important test, Unsuppressing most of the features.
I mainly changed configurations to their unsimplified counterparts. This means that both parts and mates were unsuppressed, but there were still the same number of configurations used by the assemblies that werent suppressed.
Hard Drive: 10.75 s
Network: 1m 16 s
I don't understand why exactly, but the unsuppressed version loads faster on the hard drive that the suppressed version, I suppose this shows the insignificance of suppressed items when loading a file.
This does not take into account how slow your computer is when working on an assembly.
The only thing I notice is that your assembly with most of the features unsuppressed has less resolved documents (82) than the regular assembly (93). I have no idea what a document is, as opposed to a component. Nor do I know what a Part Document is, as opposed to a Part. It's all a mystery to me.
Thanks for running the test! It certainly shows the importance of working locally and of simplifying assemblies in any way possible.