Failure to send to a routing node on another machine after limited (known) attempts

[Fraser999]

Migrated from maidsafe-archive/MaidSafe#93

Steps to reproduce:

1. Set-up 15 routing nodes on two Windows machines.
2. One has 10 nodes and the rest on the other.
3. set datasize = 500 Bytes,
4. Run "senddirect xx 500" command to send multiple messages form a routing node on one machine to node in another machine.

It seems that all the time the 48th message fails to reach to destination, and after that all messages sent to the same destination also fail to reach.
Changing the datasize seems to be related to the first failing message, however, it does not look like a linear relationship.
The same thing happens on Linux on iteration #67 when source is my system and on iteration #260 when the source is high spec 192.168.0.109 system.
Seems that sending buffer at source gets full.