Comment 1 for bug 402114

So we don't get away from having to read the index files and that repository has 18 pack files, for 18*5=90 index files that we have to probe to find revisions that we are looking for.
(I would expect 'bzr pack' to make a fairly significant difference on http download times from that repository)

Going further, the group fetch code doesn't buffer/request more than one group at a time. This could certainly be improved, though we want to avoid the place we were at before, where we tried to buffer everything at once.

Though according to some other reports we may be buffering more than we realize given the memory consumption numbers.