Yah, got it

[Jason Molenda]

Every time a message is added to an archive with mhonarc 2.5.0, an
extra copy of authors.html and subjects.html are emitted.  After
six messages were added to an archive, here's the @OtherIdx line
in the .mhonarc.db:

@OtherIdxs=(
'/sourceware/infra/ml-archiving/author.mrc',
'/sourceware/infra/ml-archiving/subject.mrc',
'/sourceware/infra/ml-archiving/author.mrc',
'/sourceware/infra/ml-archiving/subject.mrc',
'/sourceware/infra/ml-archiving/author.mrc',
'/sourceware/infra/ml-archiving/subject.mrc',
'/sourceware/infra/ml-archiving/author.mrc',
'/sourceware/infra/ml-archiving/subject.mrc',
'/sourceware/infra/ml-archiving/author.mrc',
'/sourceware/infra/ml-archiving/subject.mrc',
'/sourceware/infra/ml-archiving/author.mrc',
'/sourceware/infra/ml-archiving/subject.mrc',
'/sourceware/infra/ml-archiving/date-lowvolume.mrc',
'/sourceware/infra/ml-archiving/author.mrc',
'/sourceware/infra/ml-archiving/subject.mrc',
);

At this point, when a new message is put into the directory, mhonarc
will generate the index.html once and the authors/subjects.html
files six times.  And the next message will result in authors/subjects
being written seven times.

cygwin and gcc-patches probably stood out because they take have
long indexes and their traffic volume was high enough to make this
list long quickly.  Ah, and for the case of date-highvolume.mrc,
that index was being emitted every time as well.

In fact, I see that gcc-patches was writing its indexes 67 times
and 58 times when I finally pulled the plug.

I'm continuing to look into what's happening.  I'm willing to
believe it is a configuration mistake that I made which happened
to be OK up through 2.4.9, but invalid in 2.5.0.

Jason