In the ruby client to node server "timeout_on_sleeping_server" interop test, the client sometimes receives a Cancelled status instead of DeadlineExceeded status.

This happens frequently on jenkins interop tests. (one log in https://grpc-testing.appspot.com/job/gRPC_interop_pull_requests/9915/testReport/junit/(root)/tests/cloud_to_cloud_ruby_node_server_timeout_on_sleeping_server/)

It seems related to #53

So far this seems to only be reproducible when running interop tests similarly to how they're run on jenkins, e.g.: tools/run_tests/run_interop_tests.py -l ruby -s node --cloud_to_prod --cloud_to_prod_auth --use_docker --http2_interop -t -j 12