I was recently trying get to a problem with
multiple acks getting sent in git9, and noticed
you'd run into the same problem:

libexec/got-fetch-pack/got-fetch-pack.c:659

	/*
	 * Git server responds with ACK after 'done'
	 * even though multi_ack is disabled?!?
	 */

I think I have a theory for this, though the
misbehavior is still puzzling.

It seems that there's one ack sent *per duplicate
hash*; if all "have" lines are unique, it seems
that only one ACK is sent back.

currently experimenting on my end, let me know
if this seems to solve the problem.

(otherwise, I'm also looking at doing mutli-ack
support; it seems like it'll be easy enough)