Patent law is not always the most intuitive thing and it works in ways that tech people usually do not readily comprehend because to them many things are 'obvious' that are not obvious to a patent examiner or someone 'skilled in the art'.
And many times the software people are right (especially when it comes to software), and just as many times they are wrong or simply confused by the jargon surrounding patents.
The 'best mode' is not a requirement for a patent at all, you either have something patentable, or you do not. If it stands the test of time (other paying their license fees, you winning court cases against parties infringing) then it was a strong patent. If you come up with a better mode later on that's fine and need not be disclosed.
Whether or not a patent was the very best patent that you could have taken out is - as far as I know - not of any material impact to your chances of winning such cases.
But what I do know is this: in some places breaking NDA's and spreading trade secrets around is a very dumb thing to do so if Google can make that claim stick the other party has a serious extra problem.
The patents will be dealt with on their own, regardless of any 'modes'.
Ideally this should all be disclosed, including future improvements but it's been a long time since patents were used for their original intended purpose anyway, these days it is probably best to think of them as weaponized IP.
Forgive me for my ignorance, but isn't this exactly what a patent is intended for? An inventor discloses how the invention works, and is given a monopoly in exchange. If the patent doesn't give the details of how something works, then why should society pay the inventor for it by granting a monopoly over it?
Yes, in theory all the (critical) details are there but in practice this only works in simpler, for instance mostly mechanical patents.
Patents that are broader, for instance to patent some principle may give an illustration and enough info that if you're skilled in the art that you may re-implement it but you're definitely not going to get working schematics and blueprints to work off.
The situation is a bit better than with most scientific papers but not that much better.
[0]: http://assets.sbnation.com/assets/1701443/USD670286S1.pdf
Disclosing the best mode in the specification of an application is a requirement, pre-AIA and now: see MPEP 2165.
So, since 2012 this is now a requirement.
I've found two examples of this successfully used in litigation, they are referenced here:
https://www.uspto.gov/web/offices/pac/mpep/s2165.html
Bottom of the page, neither of those come close to Google not disclosing certain parameters.
It was a requirement before the Leahy-Smith America Invents Act, which is what is meant by "pre-AIA". You stated that "'best mode' is not a requirement for a patent at all" which was incorrect pre-AIA and remains incorrect in the AIA era. Disclosure of best mode remains a statutory requirement, even though it now lacks enforcement.
Your citation is of MPEP 2165, which is what I had referenced earlier. Thank you for sharing a direct link for others to review.
My point was that if their trade secret is also the best mode for the patent, then seemingly that puts them at risk for violating the best mode requirement and having their patents revoked, depending on when they were filed.