Since SHA-512 has stood the test of time, the argument Schneier is making is essentially:
- There is nothing wrong with SHA-512, it has not proven insecure (even though it looked like it may be 7 years ago)
- All the new systems proposed as SHA-3 are at best only very slightly smaller or faster than SHA-512, and that speed up doesn't justify the risk of pushing out something new and relatively untested (wrt a decade or so of SHA-512)
New systems are a pretty big deal, since they often have implications for the next round of hardware, recommendations that inform audits and compliance testing, etc. - so there needs to be a pretty high bar to clear between major versions of a major standard.