haproxy-1.6.2 memory consumption problem with rabbitmq (no replies)

Hi,

Haproxy-1.5.8 has been upgraded to 1.6.2 version and I have a problem with
a rabbitmq now.

RabbitMQ version 3.5.6
Debian 8 Jessie
The working config for 1.5.8 (no timeouts):

global
chroot /var/lib/haproxy
daemon
maxconn 4000
pidfile /var/run/haproxy.pid
stats socket /var/lib/haproxy/stats
user haproxy
group haproxy

defaults
log global
maxconn 2048
option tcplog
retries 3

listen rabbitmq_frontend
bind *:5672
mode tcp
option tcplog
default_backend rabbitmq_backend

backend rabbitmq_backend
balance roundrobin
mode tcp
server test-tda-rmq-01 10.1.112.126:5672 maxconn 4000 check inter 5s rise
2 fall 3
server test-tda-rmq-02 10.1.112.127:5672 maxconn 4000 check inter 5s rise
2 fall 3

For 1.6.2 - after a couple of minutes after start, the rabbitmq begin to
eat memory for every connection (about 80 connections totally) and died
soon.

Any ideas?

Regards,
Juri