I have a 1.4gb SVN repo (working copy checkout, so more like 700mb actual files) with 15k revisions. Horror of horrors, I pulled up the first commit message and look what message I found:
commit ...
Author: (no author) (no author)@...
Date: Sun Jun 8 ... 2003 +0000
Standard project directories initialized by cvs2svn.
git-svn-id: ...@1 ...
If you've got more than 15k revisions, larger than 1.4gb working directory, or a codebase that is over 10 years old then keep talking. Otherwise get over yourself and use git-svn.
170k revisions and several repositories totalling half a TB, with some working directories in the high double digit GBs. Given that we have no performance or reliability problems, and our server isn't particularly beefy - just problems with branching - I wouldn't be surprised if people have an order of magnitude larger repos.
The KDE Subversion repository had modules with upwards of a million revisions. Most have been moved to git, and had to be split into much more granular git modules (though technical concerns were not the only reason that was done, we wanted more modularity anyways).