Earlier quoted context omitted.
TCP only gives you 16 bits of checksum, which is really not enough. I used to work for a CDN which served downloads over HTTP, and when the object size grew to a few gigabytes a non-negligible percentage of users ended up with corruption. The SSL MAC is valuable even in the absence of enemy action.
This is a really interesting 'real world' anecdote. Do you know of any related data available publicly? i.e. "on our network, X bits in Y terabytes end up with undetected corruption, meaning that approximately Z% of downloads of a 2GB file will have at least one bit error".
In "Performance of Checksums and CRCs over Real Data" Stone and Partridge estimated that between 1 in 16 million and 1 in 10 billion TCP segments will have corrupt data and a correct TCP checksum. This estimate is based on their analysis of TCP segments with invalid checksums taken from several very different types of networks. The wide range of the estimate reflects the wide range of traffic patterns and hardware in those networks. One in 10 billion sounds like a lot until you realize that 10 billion maximum length Ethernet frames (1526 bytes including Ethernet Preamble) can be sent in a little over 33.91 hours on a gigabit network (10 * 10^9 * 1526 * 8 / 10^9 / 60 / 60 = 33.91 hours), or about 26 days over a T3.