We have a windows machine using 2 NIC connected to a cisco switch, the idea is somewhat double avaiable bandwidth.

With this configuration we have unexpected packet lost occuring in the switch.  Our IT guy changed the configuration 
from IEEE Load Aggreagation to Team ABL (I'm just repeating what I'm told). This has greatly improve the situation, 
however with TEAM ABL sloginfo is full of

May 05 18:11:06    5    14     0 arp info overwritten for 10.121.205.100 by 00:1b:21:9f:0d:96
May 05 18:11:38    5    14     0 arp info overwritten for 10.121.205.100 by 00:1b:21:9f:0d:97

I've always been tough that having two NICs using the same IP adress is really bad.  Yet this seems to be working.

Does it work by sheer luck or by design?