The current implementation of fetchindex removes the bz2 file when it generates the INDEXFILE. This defeats the purpose of -m flag (mirror mode) for fetch.
The included implementation separates the fetch from the unpacking. The unpacking is done in a way which doesn't remove the bz2 file. This way multiple fetchindex runs without any change to the master INDEXFILE on the website doesn't result in downloads. This should reduce the traffic and the time needed for the user.
Additionally this also respects FETCH_ENV for the FETCHINDEX command.
Over to maintainer.
Adding manually to audit trail:
Please see the combined patch in ports/89809 which attempts to fix
multiple problems in this area.