how man replication paths for vsa on vsphere are used

[Christoph Schlagintweit]

first of all a description of the used configuration:

2x HP DL370G6
each configured with 1 P410i SMA controller (512MB Battery Cache)
each configured with 16 SFF HDD 146GB/10K RAID10 config.
both are running esx4u1 - all current updates
i've got one VSA running on Server1 the other on Server2 (on both hosts at local vmfs datastore)
my FOM (Failover Manager VM) is running on a physical/seperated host (running win2008)

i've configured the lefthand cluster in the following way:
both vsa are in the same management group and in the same (standard) cluster
i've created two testing luns one 80GB - the other 600GB
both are replicating 2way

on the esx side i'm using standard nics to run the iscsi traffic (on every server i've configured 4 physical nics for it)
i've enabled the "software initiator multipathing" using round robin policy in esx (new feature in esx4), jumbo frames is also configured on all components.
basically this seems to work - in the storage config tab on my esx hosts i'm seeing that there are 4 paths to every lun (from the vsa)
and policy is round robin.

so finaly heres my problem (if it ist?)
replication of the volumes from vsa1 to vsa2 seems to use only one nic port and this looks like a bottleneck to me.
for example when i started my test i'Ve started with only vsa1 in the cluster - so no replication at all. i've seen a transfer rate up to 375MB/s but
at the moment i've added vsa2 to the cluster - activated 2-way replication and waited that the resync has finished - my transfer rate went down to 120MB/s
with the same workload test.

has anyone experienced the same behaviour when replication is on?