qa/1024 and disk partitions/docker indom issues

["Ken McDonell" <kenj@xxxxxxxxxxxxxxxx>]

qa/1024 (and other similar tests) are failing on ALL my QA hosts, apparently
due to non-determinism in the assignment of instance ids to instances.

disk.partitons.* seems particularly afflicted  ... also the instance id for
the "docker ..." instance of disk.dm.*.

All of the data is there, it is just the internal instance ids that are not
deterministic.

Is this expected?

I have a fix for 1024 that maps the instance ids to the constant "N" and
sorts the list of instances in each block of pminfo output ... I just want
to check this is expected before I commit.