The whole history will be lost when following this procedure, which renders
it inappropriate for backup repositories whose history actually exists.
However, if I get Tim right, he is talking about an interrupted _initial_
backup run. In this case, giving up the entire history obviously does not do
any harm. Without any metadata (i.e. the rdiff-backup-data subfolder), afaik
a forced rdiff-backup run will take whatever there is in the target folder
as a base and won't retransfer existing parts. So my solution might be the
right thing to do in this special case.