Hello Mirza,
EIGRP has its own algorithm for managing the process of sending updates to neighbors.
This reliable transport protocol allow to send first try as a multicast.
EIGRP keeps a state and a queue for each neighbor.
If a neighbor doesn't ack the reception of the multicast update within a time, the EIGRP process takes note of this.
Further attempts to send a copy of that update are sent with destination unicast = neighbor ip address.
The device attempts some times to resend the update if all times no ack is received there is a max number of retries.
After max retries expires the EIGRP neighborship is reset because the neighbor is considered not reliable.
Edit:
you can be facing an MTU issue here that causes a specific update of a certain size to be dropped in between the two devices.
if possible you should check on the neighbor if the update packet arrives.
Actually, update size depends on the prefix length of carried routes.
routes with prefix len > 24 takes 4 more bytes to be carried.
This has been explained in a thread of some mounths ago.
Hope to help
Giuseppe
