Largely varying copy speeds

Deploying a single installation to identical computers on the same switch seems to be varying its copy speed in a large difference:

EX in attached image:

Deploying to 17 identical PC's that are all in one lab on the same switch....

Deployed all simultaneously, most are doing around 2-4 MB/sec but some get stuck at 150KB/sec and below and never increase.  From start until the time I killed them (gave them 15 minutes+) they remained at their slow speed in comparison.

 When re-deploying to the same machines (say 5 were aborted here due to their slow speeds) those ones that had slow speeds previously were as fast as the other ones were on the first deployment (2-4MB/sec) except again a few were stuck (this time 2).  Eventually the same thing happened, I aborted and re-deployed to those remaining two and then their speed was normalized.

It seems very strange that their speeds are fluctuating so much, some going literally 20x slower than other that are identical and from the same install and then right after aborting will be normal as if there was never a problem.  I have tried bandwidth at different %'s to limit throttling (even 100%) and this still happens.  I have also tried using pull method vs. push and it is even more varying.

Anyone have any info as to why this may be happening and how to prevent it? I can re-create it almost every time I deploy to 10+ machines that are identical.