the notion think civil discovery against private telecoms is a good avenue to find out what the government was doing is both stupid and abusive litigation.  Sue the government directly or lobby Congress to investigate.

There are plenty of ways to obtain discovery.  Hell, the state secrets privilege is common law; Congress could partially repeal that themselves.  Also, JJE noted several other, more politically palatable methods.  

In any event, the issues you identified are not constitutional in scope.

I suppose it's probably little consolation if you presume that restrictions in the bill are only cosmetic, but the bill specifically says that no one may be targeted under this act solely for activities protected by the First Amendment.  Specifically, you can't use protected speech as probable cause for a FISA warrant.

I'll pose it from the other direction: why would the NSA be going after backbone snooping if they didn't think they could get something useful out of it?

I think you overestimate the strength of consumer-grade encryption, or underestimate the ability of the NSA to crack it.

The most widespread form of encryption, Public-key cryptography, used by skype and web browsers, has very little overhead.  It works by multiplying large prime numbers together, which for computers takes a negligible amount of resources (e.g. an everyday  2Ghz 64-bit dual core could perform two billion 128-bit multiplications per second).  Multiplying is easy, but factoring a number with only two prime factors is extremely hard, even for modestly sized numbers because of the exponential growth of the problem.  We can multiply a billion bit numbers effortlessly but we can't factor 100 bit semiprimes.  Anyways, a 20-year old computer could easily cope with 128 bit SSL used in browsers, even if it couldn't render the page.

Because of this and other things, the term 'consumer-grade encryption' is somewhat meaningless these days, especially in the era of open source software.  Very few algorithms used in encryption today are proprietary, so anyone can implement them and share that implementation with everyone.  In fact, the methods that are freely available for consumers to inspect the guts of are widely considered to be superior to closed-source version as the openness of these systems means that the world's top cryptography experts can and do continually vet them for errors.  There's immense prestige afforded to the researcher that breaks a high-profile cipher, and the number of such people working worldwide dwarfs the resources of the NSA.  This means that the public has cost-free access to top of the line security products that have been mathematically proven to be unbreakable unless the attacker is in possession of nearly infinite computing resources.

(for example, any 256-bit algorithm has a keyspace greater than the number of atoms in the universe.  If our computers were as efficient as information theory allows, using all the matter on Earth to construct them, and we had 100% efficient access to all  the remaining power in our sun, we still wouldn't be likely to brute-force a 128-bit key.)

You're right that even in the recent past it's been possible to brute force commonly used encryption.  The Electronic Frontier Foundation made big waves in 1998 by building their own massively parallel dedicated cracking engine called Deep Crack for under $250,000 that was able to try every possible combination in a message encrypted with the then US government standard DES.  It finished working in 56 hours.  By the next year, using distributed online resources, that time was halved.  Clearly the NSA has more than ample resources to duplicate this feat.  But DES was widely criticized at the time for having a 56-bit keyspace.  Adding one bit doubles the search space, so even a hypothetical 60-bit DES would take 16x more resources to crack by this method.  With this exponential growth, by the time you hit 128 and 256 bits, you've made brute-force attacks impossible for any government.  

Of course just having 128-bits isn't a guarantee that your method is strong, the 802.11's WEP can utilize a 128-bit key and still be broken in seconds on an active access point using a replay attack (useful in a pinch).  This was replaced by WPA which is as far as we know unbreakable.  This pattern of encryption being broken and replaced has been ongoing, but it's not going to last forever.  In the future, like WEP, some of our methods of encryption will be shown to have weaknesses that allow governments to sidestep their mathematical soundness, but the whole world has access to these methods and can find these attacks too.

At some point soon, if we haven't reached it yet, the only feasible method of breaking encryption will be monitoring the source or sink of the data, which of course can't ever be completely defended against.  If, as I predict, encryption becomes extremely widespread for mundane communication (and there's no reason to believe it won't be -- it's prudent and the trend is obvious), snooping at the routers will be useless.

My CS background probably makes me overconfident in computers' abilities, but nothing I've said above is false.  

""warrantless wiretapping"

they weren't wiretapping, they were tracking who was talking to whom, not necessarily what they were saying.

It was all about patterns...

I do think the NSA will likely continue to break the law, as long as they and telecoms are being provided legal cover.  

If this is true, pretty much the only thing Senator Obama can do to fix it is become President Obama :)