It's interesting for me how spy bonuses affect the whole technology stealing process. I saw that such an article before but there was no clear answer about it. Now in VDC mode we have rather cheap "price" for +20 spying.
You're right, so far there is no clear answer on spying bonuses efficiency and practice shows contradictory results: sometimes you steal with fewer spies and worse bonuses, but perhaps it's just lucky dice rolls.
I will answer your first question now, as for the second maybe I will have clear answer later (tf8 and I are working to decode the spying formula).
What can you say (or refer to the corresponding topic) about spying efficiency in pure 1vs1 or with 2 AIs?
The key in spying efficiency is time. MP games are very much time limited, after the contact you
usually have no more than 20-30 turns, then follows dénouement...during this time you can steal techs. Sometimes stealing a key tech like ship computer, armor or VRN (or Supercomputers)
can decide the whole game. Naturally the chance of success of stealing is determined by the spying bonuses: the better spying bonuses the more chances. However, what tech you will steal is also random. You may steal couple of techs, but not so important ones. So efficiency is doubly random.
The serious breakthrough is when you begin to steal techs one by one, which happens when your offense spying bonus much prevails over opponent's defense bonus. But this depends on opponent's race. It may happen that you take +20 spying bonus, while your oponent takes unification and all your spying strategy is lost in vain, you lose picks just so and your race is slower without any gain. For instance, with the same cost you could take warlord, which would be of much more benefit...As a conclusion: +20 spying bonus could be effective against demo and dict with -spy negative, in other cases it's a zilch. As one of my friends said: "your race would devastate demo, but now (vs uni) it's doom and gloom"