Redis 4.0.1, Slave blocked from reaching master node

[ghost]

Hi,
I have a problem with Redis 4.0.1 with classical cluster configuration.
First and most important thing to tell you is that I run cluster without any problem on existing configuration, when I use Redis 3.2.10.
However under Redis 4.0.1 my Redis cluster is highly unstable, and often many slave nodes permanently can't join cluster.
On one of this slaves, through log files I managed to catch bug, that I will try to post here in best possible way.
Again before further text I need to say one more time, that I am using same hardware and OS configuration without any problem with Redis 3.2.10
So here we go.
Cluster Architecture:
90 nodes on 3 machines where we have 30 master nodes and 60 slave nodes. Each master node has 2 slave nodes.
Master and 2 slave nodes are always distributed across 3 machines
Hardware configuration:
I use AWS. Here are 3 R4 instances 4xlarge. Each machine has 16 cores and 122 GB RAM, with 200GB of General Purpose SSD, and static IP addresses
OS configuration:
Amazon Linux AMI 2017.03.1 HVM,
OS parameters
sysctl -w net.core.somaxconn=65535 inside /etc/rc.local
echo 'vm.overcommit_memory = 1' >> /etc/sysctl.conf
echo never > /sys/kernel/mm/transparent_hugepage/enabled inside /etc/rc.local
fs.file-max = 100000 inside /etc/sysctl.conf
<my_user> soft nofile 4096 inside /etc/security/limits.conf
<my_user> hard nofile 10240 inside /etc/security/limits.conf
session required pam_limits.so inside /etc/pam.d/

Bug dump is attached
RedisBugDump.txt