memcached restarting during xdcr tests

While running this test:
python testrunner -i cheri.ini -t xdcr.uniXDCR.unidirectional.load_with_failover,replicas=1,items=10000,ctopology=chain,rdirection=unidirection,doc-ops=update-delete,failover=source

I'm seeing memcached being started and then a CRASH REPORT:

ns_server:debug,2014-01-16T10:16:46.512,ns_1@172.23.106.55:ns_memcached-default<0.10082.2>:ns_memcached:init:141]Starting ns_memcached
[ns_server:debug,2014-01-16T10:16:46.513,ns_1@172.23.106.55:<0.10083.2>:ns_memcached:run_connect_phase:164]Started 'connecting' phase of ns_memcached-default. Parent is <0.10082.2>
[error_logger:error,2014-01-16T10:16:46.513,ns_1@172.23.106.55:error_logger<0.6.0>:ale_error_logger_handler:log_report:72]
=========================CRASH REPORT=========================
crasher:
initial call: erlang:apply/2
pid: <0.10076.2>
registered_name: []
exception error: no function clause matching
mc_client_binary:rev_to_mcd_ext(

{2, <<0,11,66,150,67,116, 111,104,0,0,0,0,0,0, 0,0,1,0>>}

)
in function mc_client_binary:do_update_with_rev/7
in call from ns_memcached:do_handle_call/3
in call from ns_memcached:worker_loop/3
ancestors: ['ns_memcached-default','single_bucket_sup-default',
<0.9876.2>]
messages: []
links: <0.10073.2>,#Port<0.52126>
dictionary: []
trap_exit: false
status: running
heap_size: 28657
stack_size: 24
reductions: 2792
neighbours:

cbcollect: 10.1.1.210:7878/cherimoya-s21603.zip