Projector consuming large amount of CPU for tiny workload with no indexes defined

Created a cluster with three nodes (node1 & node2 KV, node3 index & query), default bucket with 512MB bucket quota.

I then started a modest workload (1000 documents 100bytes each, 100 op/sec) using pillowfight:

cbc-pillowfight -U couchbase://192.168.46.101/default --rate-limit 100 -B 1

During this, I looked at the CPU usage on node1. I see the projector consuming 30% CPU:

  PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND

 8956 couchbas  20   0  498m  56m 6968 S 30.5  1.5   3:24.23 projector

 8735 couchbas  20   0  472m 144m 6992 S 15.9  3.8   3:30.58 memcached

Note there are /no/ indexes defined on the default bucket (be them mapReduce or GSI). Why is the projector using /any/ CPU at all??