I have a test environment with two 150GB discs, configured as a ZFS mirror (150GB usable space). I did a AVS configuration to replicate that discs to another server. I have lost the sync between the servers, and the state on the primary node was not sane.. so, i did a full resync from the secondary server.
Here is the time waste to make that full resynchronization (secondary -> primary).

# dsstat

name                   t  s    pct    role   ckps   dkps   tps  svt
dev/rdsk/c2d0s0   P RS  99.89  net      -       Inf     0  -NaN
dev/rdsk/c2d0s1                    bmp      0      102    0  -NaN
dev/rdsk/c3d0s0   P RS  99.89  net      -       Inf     0  -NaN
dev/rdsk/c3d0s1                    bmp      0       73     0 -NaN

# time /usr/sbin/sndradm -C local -g POOLNAME -n -w

real    464m18.109s
user    0m0.107s
sys     0m0.075s

Almost 8hs…

# dsstat

name                  t  s    pct   role   ckps   dkps   tps  svt
dev/rdsk/c2d0s0   P  R   0.00  net       -      0       0    0
dev/rdsk/c2d0s1                   bmp      0      0       0    0
dev/rdsk/c3d0s0   P  R   0.00  net       -      0       0    0
dev/rdsk/c3d0s1                  bmp      0       0       0    0

The interface is a Gigabit ethernet, and is the same interface used to access another services. I think a good approuch is have a dedicated interface to handle the sync/resync tasks. In this specific case, i don’t know if the time would change much, because the server was idle, and all the traffic was basically the AVS one…