It seems TCPConnectionMap didn't restore after network failure

I got next exception on node (let say node1).

WARN [ConnectionMap.Acceptor [xxx.xxx.xxx.xxx:34383],null,null] org.jgroups.protocols.TCP [JGRP00006] failed accepting connection from
peer: %s
java.net.SocketTimeoutException: Read timed out
at java.net.SocketInputStream.socketRead0(Native Method) ~[na:1.7.0_17]
at java.net.SocketInputStream.read(SocketInputStream.java:150) ~[na:1.7.0_17]
at java.net.SocketInputStream.read(SocketInputStream.java:121) ~[na:1.7.0_17]
at java.io.BufferedInputStream.fill(BufferedInputStream.java:235) ~[na:1.7.0_17]
at java.io.BufferedInputStream.read1(BufferedInputStream.java:275) ~[na:1.7.0_17]
at java.io.BufferedInputStream.read(BufferedInputStream.java:334) ~[na:1.7.0_17]
at java.io.DataInputStream.readFully(DataInputStream.java:195) ~[na:1.7.0_17]
at org.jgroups.blocks.TCPConnectionMap$TCPConnection.readPeerAddress(TCPConnectionMap.java:495)
at org.jgroups.blocks.TCPConnectionMap$TCPConnection.<init>(TCPConnectionMap.java:377)
at org.jgroups.blocks.TCPConnectionMap$Acceptor.handleAccept(TCPConnectionMap.java:299)
at org.jgroups.blocks.TCPConnectionMap$Acceptor.run(TCPConnectionMap.java:283)
at java.lang.Thread.run(Thread.java:722) [na:1.7.0_17]

After it two nodes works in next way:

node 1 - sends Discovery requests every 3 seconds:
[2013-08-05 21:02:00,585] TRACE [TransferQueueBundler,global,_index-subscriber-node01] org.jgroups.protocols.TCPPING _index-subscriber-node01: sending discovery request to xxx.xxx.xxx.xxx:34383

node 2 - [2013-08-05 21:02:03,791] TRACE [OOB-2,global,_index-subscriber-node02] org.jgroups.protocols.TCPPING _index-subscriber-node02: received GET_MBRS_REQ from _index-subscriber-node01, sending response [PING: type=GET_MBRS_RSP, arg=_index-subscriber-node02, view_id=[_index-subscriber-node03|230], is_server=true, is_coord=false, logical_name=_index-subscriber-node02, physical_addrs=xxx.xxx.xxx.xxx:34383]

And node 1 - didn't get any response and continue to send discovery request every 3 seconds.

So it necessary to restart node to restore functionality.

What is interresting - I see much more simmilar exceptions - and in most cases functionality is restored authomatically. Only few of them break a node.