In your Cisco EIGRP network, you notice that the neighbor relationship between two of your routers was recently restarted. Which two of these choices could have made this occur? (Choose two.)

A.
An update packet with init flag set from a known, already established neighbor relationship was received by one of the routers.

B.
The ARP cache was cleared.

C.
The counters were cleared.

D.
The IP EIGRP neighbor relationship was cleared manually.

Explanation:
The following are the most common causes of problems with EIGRP neighbor relationships:

Unidirectional link
Uncommon subnet, primary, and secondary address mismatch
Mismatched masks
K value mismatches
Mismatched AS numbers
Stuck in active
Layer 2 problem
Access list denying multicast packets
Manual change (summary router, metric change, route filter)

According til Ivan Pepelnjak’s book "EIGRP Network Design Solutions" the Init flag is set in the initial update packet when to neighbors discover each other and start their initial topology table exchange.

There are two basic purposes for the init flag. First, it’s a part of the three way handshake that eigrp uses when building an adjacency.

— Router B comes up on a wire.
— Router A receives Router B’s hello, and places it in "pending" state. This is a not completely formed adjacency; as long as B is in this state, A won’t send any routing information to it.
— Router A sends an empty unicast update with the init bit set.
— Router B receives this update with the init bit set, and places Router A in the "pending" state.
— Router B now transmits an empty update with the init bit set, unicast, to A. This empty update also contains the acknowledgement for Router A’s init update (that this ack is piggybacked is an integral part of the three way handshake process).
— Router A, on receiving this init update, places Router B in the "neighbor" state, and sends an acknowledgement for the init update from Router B.
— Router B receives this ack, and places A in "neighbor" state.
The two routers can now exchange routing information, knowing they have full two way connectivity between them.

The second use of the init bit is more esoteric. Suppose you have Routers A and B, running along fine, for many hours. Router A reloads, but comes back up before Router B’s hold timer has expired. When Router B sees A’s hellos, it will assume that A just missed a couple, and everything is fine. But everything isn’t fine–A just lost all of its routing information! How can A signal this state, and as B to resynchronize?
A can send an empty update, with the init bit set. This causes Router B to place A in the "pending" state, and wipe out all the information it’s learned from A (unless, of course, graceful restart is configured/etc).