Lots and lots of algorithms gain an extra lg(n) factor if you drop that assumption.
e.g. http://blog.computationalcomplexity.org/2009/05/shaving-logs...