The current way the subversion loader handles incremental loading is not optimal at all. Indeed we need to fully dump a repository to disk for each visit regardless of what we have loaded so far (due to limitations of the svnrdump tool).

Fortunately, there is another subversion dump tool called rsvndump without the svnrdump limitations that could allow us to only dump to file the revisions not loaded so far into the archive and properly resume loading.

If this works, incremental loading of big subversion repositories would be much more efficient so we should investigate that approach.