Response time spikes when connections increase

We have clients that connect using 300-400 connections parallely and we have a src_conn_cur reject set to only if client hits 1000 connections.
When ever we have the clients connect during that period we see there is a spike in response time for approximately 1-2% of messages.
Is this common? config

global
chroot /var/lib/haproxy
daemon
log 127.0.0.1:12000 local0
maxcompcpuusage 90
maxcomprate 1000000
maxconn 340000
pidfile /var/run/haproxy.pid
stats socket /var/run/haproxy.sock mode 777 level admin
tune.ssl.cachesize 1000000
tune.ssl.default-dh-param 2048

defaults
log global
mode http
option httplog
option redispatch
option http-no-delay
retries 3
timeout connect 5000
timeout server 1800000
timeout client 1800000

frontend XXXXX
bind 10.XXXX:443 ssl crt /etc/pki/tls/certs/XXX
mode http
default_backend bk_xms
maxconn 340000
option forwardfor header X-Real-IP
option httplog
option http-ignore-probes
stick-table type ip size 100k expire 30s store conn_cur
tcp-request connection reject if { src_conn_cur ge 1000 }
tcp-request connection track-sc1 src

backend XXX
mode http
balance roundrobin
http-check expect status 200
http-reuse aggressive
option allbackups
server XXXXX1 10.xx.xx.xx:8080 check fall 3 rise 5 inter 2000 weight 10
server XXXXX2 10.xx.xx.xx:8080 check fall 3 rise 5 inter 2000 weight 10

Need urgent help

Hello,

Are you sure Haproxy is responsible for the increase of response time. In the log, you can see the time repartition between connection, requests, backend response time, … what do the timers look like in normal time and during the “outage” ?

Regards,