Waiting for Agent - Exhausted Connections?

So I'm wondering if anyone had seen this condition....

Here is the issue- if I deploy a package to about 25 machines I see about 5 active "threads" deploying at the same time, all using the agent showing "internal". If I abort the 5 active deployment threads (because they were "stuck"), the entire deployment just spins showing "Waiting for Agent" or Queued with no more deployments- leaving me the only choice to abort the entire deployment and then restart the services.

This is how I got here:

I'm running remote robocopies in a deployment package- it generally works, but I've run into a few issues that seem to lock up the deployment on a target for longer than say 5 mins... Now this isn't the issue this question is about- I'm fixing those issues... BUT I see the deployment via an agent is "stuck" because of my script, so I abort it.

Now I normally see about 5 machines getting a simultaneous deployment, but when I abort one, this seems to reduce the number of simultaneous deployments by one... Now only 4 are running at the same time... kill another, 3 are running... if I kill the other 3- the entire deployment sits there saying its deploying, but everything is either "waiting for agent" or queued and it can sit there for hours- nothing, and no deployments running in PDQ (according to the bottom).

Now what's interesting is if I abort the entire deployment, not particular machines it will stop. But if I then select machines that "failed-aborted" and say deploy them once- individually- they will forever be "waiting for agent" still. BTW, if I try to deploy to a machine not in that original deployment target list- it works just fine... just those people in aborted "waiting for agent" list are stuck getting no updates- forever- or until I recycle the deploy and inventory services... then they are back in business- everything is back in business.

Also- I tried to set the timeout on the deployment to manage this. Now this didn't work very well because the timeout is at the deployment level, not each target- so when I set it to 3 mins- well at 3 mins the normal 20 machines waiting all got killed from timeout (this would be more handy IMO if the timeout was per target too). I thought maybe I could work around this by creating a larger "bulk" deployment with a default timeout and then nesting the other timeout as step 1- Nope.

Anyway, going back to the beginning, the net result is that if I deploy to about 25 machines I see about 5 active "thread" deploying at the same time. If I abort the 5 "stuck" active deployment threads in a deployment group- using the "internal" agent, the entire deployment just spins with no more deployments- leaving me the only choice to abort the entire deployment and then restart the services.

Have you seen this? Does the 5 simultaneous agent deployments seem to make sense?

Thanks in advance for the help.