The non-default configuration options used by the application:

readPreference: nearest
cluster:
  hosts: mongodb-oracle-a1-ci-set1, mongodb-oracle-b1-ci-set1, mongodb-oracle-c1-ci-set1
codeRegistry: jackson (https://github.com/ylemoigne/mongo-jackson-codec)

We have a replica set with 3 members, identified by host names mongodb-oracle-a1-ci-set1, mongodb-oracle-b1-ci-set1, and mongodb-oracle-c1-ci-set1. There is a Java application running on a fourth host. If I stop any one of the mongod daemons, primary or secondary, while the application is running, CPU for the application process goes to 100% and the application is unusable until restarted. I can replicate the issue consistently.

Attached is the log output when replicating the issue.

1. The application is started with mongodb-oracle-a1-ci-set1 daemon stopped. Everything looks fine. The other two daemons are operating normally.
2. At about 20:24:23, I start mongod on mongodb-oracle-a1-ci-set1. Everything looks fine. The new member joins the replica set and the log indicates that the application connects to it.
3. At about 20:25:05, I stop mongod on mongodb-oracle-a1-ci-set1. The application process CPU goes to 100%. From the logs, it looks like the application still thinks its connected to the missing member.
   4. At about 20:26:04, I start the mongod on mongodb-oracle-a1-ci-set1. The application process CPU remains at 100%.

Note that no mongo queries are executed. The application is sitting idle when all this occurs. Once in a bad state, the application is unresponsive and must be restarted.