When the system is under heavy load reading from bigchaindb event stream API is highly unreliable. This behaviour was observed while doing benchmark testing using this tool. For comparison when I updated the benchmark tool here to read directly from tendermint instead of bigchaindb event stream, the results became much more reliable.

To reproduce the issue run benchmark test with high number of requests on the network:
bigchaindb-benchmark --csv 18.csv -p http://<node-1-ip>:9984 -p http://<node-1-ip>:9984 -p http://<node-1-ip>:9984 -p http://<node-1-ip>:9984 --processes 16 send -r 100000 --mode sync