The backend doesn't exit cleanly

[sersorrel]

testing_nginx_1 exited with code 0
testing_frontend_1 exited with code 0
postgres_1  | 2019-08-29 15:50:20.855 UTC [6719] LOG:  unexpected EOF on client connection with an open transaction
testing_backend_1 exited with code 137
postgres_1  | 2019-08-29 15:50:21.458 UTC [1] LOG:  received smart shutdown request
postgres_1  | 2019-08-29 15:50:21.480 UTC [1] LOG:  worker process: logical replication launcher (PID 31) exited with exit code 1
postgres_1  | 2019-08-29 15:50:21.480 UTC [26] LOG:  shutting down
postgres_1  | 2019-08-29 15:50:21.615 UTC [1] LOG:  database system is shut down
testing_postgres_1 exited with code 0

Everything else exits properly when it's told to, but the backend has to be killed. This tends to make Postgres unhappy the next time it starts up.