CAS fails with replica CAS different

Run there automated CAS test fails:

./testrunner -i ~/servers/tunable-vms2.ini -t castest.opschangecas.OpsChangeCasTests.ops_change_cas,doc_ops=update\;delete\;expire,items=16,mutate_times=100,nodes_init=2

It fails because the replica CAS it not the same as the active. This is 100% reproducible. It is also seen on 4.5.0.

Here is a sample output:
2016-01-21 12:13:57 | INFO | MainProcess | test_thread | [opschangecas.verify_cas] Use item cas 1453407006434525184 to mutate the same item with key nosql0 successfully! Now item cas is 1453407007662145536
2016-01-21 12:13:58 | INFO | MainProcess | test_thread | [opschangecas.verify_cas] Use item cas 1453407007662145536 to mutate the same item with key nosql0 successfully! Now item cas is 1453407008920240128
2016-01-21 12:14:00 | INFO | MainProcess | test_thread | [opschangecas.verify_cas] Use item cas 1453407008920240128 to mutate the same item with key nosql0 successfully! Now item cas is 1453407010132131840
2016-01-21 12:14:01 | INFO | MainProcess | test_thread | [opschangecas.verify_cas] Use item cas 1453407010132131840 to mutate the same item with key nosql0 successfully! Now item cas is 1453407011320102912
2016-01-21 12:14:02 | INFO | MainProcess | test_thread | [opschangecas.verify_cas] Use item cas 1453407011320102912 to mutate the same item with key nosql0 successfully! Now item cas is 1453407012559257600
2016-01-21 12:14:03 | INFO | MainProcess | test_thread | [opschangecas.verify_cas] Use item cas 1453407012559257600 to mutate the same item with key nosql0 successfully! Now item cas is 1453407013821874176
2016-01-21 12:14:05 | INFO

AssertionError: replica cas mismatch. Expected 1453407015166935040, actual 1453407013821874176.

This is seen on build 1061 and I think was introduced in the last few days.