Some race between insertion and delitions items to couch_setview_name_to_sig table

Farshid got this while running test from this change set: http://review.couchbase.org/#change,12361. Among plenty of other errors in the log there is this one:

=========================CRASH REPORT=========================
crasher:
initial call: couch_event_sup:init/1
pid: <0.187.0>
registered_name: []
exception exit: {'EXIT',
{{case_clause,
[{<<"default">>,
{<<"_design/dev_test_view_on_10k_docs-c5abd37">>,
<<57,37,41,201,93,208,192,231,227,9,177,64,
251,180,145,114>>}},
{<<"default">>,
{<<"_design/dev_test_view_on_10k_docs-c5abd37">>,
<<114,187,98,215,41,151,58,48,85,66,14,103,
185,80,79,64>>}}]},
[

{couch_set_view,handle_db_event,1}

,

{couch_db_update_notifier,handle_event,2}

,

{gen_event,server_update,4}

,

{gen_event,server_notify,4},
{gen_event,server_notify,4}

,

{gen_event,handle_msg,5}

,

{proc_lib,init_p_do_apply,3}

]}}
in function gen_server:terminate/6
ancestors: [couch_set_view,couch_secondary_services,couch_server_sup,
cb_couch_sup,ns_server_cluster_sup,<0.41.0>]
messages: []
links: [<0.186.0>,<0.153.0>]
dictionary: []
trap_exit: false
status: running
heap_size: 377
stack_size: 24
reductions: 152
neighbours:

After a quick look at handle_db_event code it seems that this function does not expect to find several signatures for the same design document. So there must be some race between the code that calls add_to_ets and the one that calls delete_from_ets.