Server config advice needed

Testing one of my test applications I am observing (to my eyes) a strange behavior described in A and B:

shinycannon recording.log --workers 20 --loaded-duration-minutes 2 --output-dir apptest3_load --overwrite-output --start-interval 50

A: The number of connections raises up to its limit with two processes running. According to the server utilization_scheduler values. After a while the number of connections begin to drop or increase UNDER/OVER the expected limit causing the server to crash after a while. The consumed RAM during the test is at most a bit over 1/3. There's enough of disk space and no more than 30% of CPU.

shinycannon recording.log --workers 50 --loaded-duration-minutes 2 --output-dir apptest3_load --overwrite-output --start-interval 50

B: The number of connections raises up to its maximum, but only two R processes are running while expecting three. After a short while the server crashes.

Shiny-server pro 1.5.7
CentOS 7 x86_64
4 GB RAM on CentOS7.

I have one test application and I've set my utilization scheduler on the global level (all apps) to:
utilization_scheduler 20 .9 5;

Please advice. Thank you.