APNS connections stay open

From our user list we got a report with too many ESTABLISHED connections to APNs:
http://lists.jboss.org/pipermail/aerogear-users/2017-January/001034.html

It is working fine so far, however after some time we see plenty of open
connections to apple servers (17.188.**). Today we hit a peak with over 3000
open connections. It seems that each connection is tied to a separate
thread, thus we had over 3000 threads as well. This is when our monitoring
fired and our operation guy got alerted.

out of the users netstat:

(...)
tcp        0      0 nbgpbossba1.petaf:46382 17.188.138.28:2195      ESTABLISHED
tcp        0      0 nbgpbossba1.petaf:52186 17.188.161.118:2195     ESTABLISHED
tcp        0      0 nbgpbossba1.petaf:52203 17.188.161.118:2195     ESTABLISHED
tcp        0      0 nbgpbossba1.petaf:34515 17.188.163.204:2195     ESTABLISHED
(...)

Verification steps:

Deploy changes to eteam cluster
Switch UPS to mocked APNS
Use test-driver to generate notifications.
Monitor connections to APNS during sending notifications (there should be spike in connections).

netstat | grep 16002

Monitor connections to APNS after notifications are delivered (number of connections should go down, ideally to 0).