seeing spurious error from select_bucket during compaction

See logs in MB-10286. On node .74 I'm seeing this:

Regarding main problem I'm not sure we have much c

[ns_server:info,2014-02-23T13:07:00.497,ns_1@10.1.3.74:<0.17813.1>:compaction_daemon:maybe_compact_vbucket:928]Compaction of <<"default/229">> has finished with ok
[user:info,2014-02-23T13:07:00.498,ns_1@10.1.3.74:ns_memcached-default<0.690.0>:ns_memcached:terminate:777]Shutting down bucket "default" on 'ns_1@10.1.3.74' for deletion
[ns_server:info,2014-02-23T13:07:00.500,ns_1@10.1.3.74:<0.17823.1>:compaction_daemon:maybe_compact_vbucket:921]Compacting <<"default/230">>
[ns_server:info,2014-02-23T13:07:00.501,ns_1@10.1.3.74:<0.17823.1>:compaction_daemon:maybe_compact_vbucket:928]Compaction of <<"default/230">> has finished with {error,
{select_bucket_failed,

{memcached_error, key_enoent, <<"Engine not found">>}

}}

You can see that vbucket compaction worked fine, and then suddenly it fails.

Likely something in bucket engine.

ns_servers's code is using common (shared across all buckets) pool of connections. All auth-ed using admin credentials. And it's therefore doing select_bucket after taking connection from pool every time. AFAIK it used to work.