[FTS] queries fail under low-amount of concurrent load on multi-node cluster due to too many connections

cbft opens tcp connections between nodes but doesn't reuse them.
Amount of connections quickly reaches the limit and FTS queries start returning connection errors:

"status": {
"errors": {
"IndexClient - http://172.23.99.39:8094/api/pindex/perf_fts_index_420d95bfb1812a98_321cbb28": "Post http://%40fts-cbauth:3e9bc4ea6b573dd5fefe64150f9bc385@172.23.99.39:8094/api/pindex/perf_fts_index_420d95bfb1812a98_321cbb28/query: dial tcp 172.23.99.39:8094: connect: cannot assign requested address",
.

netstat output:

[root@fts01-s424 ~]# netstat -lpnta | grep "TIME_WAIT - " | wc -l
55774

[root@fts01-s424 ~]# netstat -lpnta
tcp 0 0 172.23.99.211:51190 172.23.99.39:8094 TIME_WAIT -
tcp 0 0 172.23.99.211:34465 172.23.99.39:8094 TIME_WAIT -
tcp 0 0 172.23.99.211:57708 172.23.99.39:8094 TIME_WAIT -
tcp 0 0 172.23.99.211:47835 172.23.99.40:8094 TIME_WAIT -
tcp 0 0 172.23.99.211:33509 172.23.99.40:8094 TIME_WAIT -
tcp 0 0 172.23.99.211:54078 172.23.99.40:8094 TIME_WAIT -
tcp 0 0 172.23.99.211:47845 172.23.99.39:8094 TIME_WAIT -
tcp 0 0 172.23.99.211:57334 172.23.99.39:8094 TIME_WAIT -
tcp 0 0 172.23.99.211:51675 172.23.99.39:8094 TIME_WAIT -
tcp 0 0 172.23.99.211:39886 172.23.99.39:8094 TIME_WAIT -
tcp 0 0 172.23.99.211:60856 172.23.99.39:8094 TIME_WAIT -
tcp 0 0 172.23.99.211:50033 172.23.99.39:8094 TIME_WAIT -
tcp 0 0 172.23.99.211:49765 172.23.99.39:8094 TIME_WAIT -
tcp 0 0 172.23.99.211:40220 172.23.99.39:8094 TIME_WAIT -
tcp 0 0 172.23.99.211:49538 172.23.99.40:8094 TIME_WAIT -
tcp 0 0 172.23.99.211:46047 172.23.99.40:8094 TIME_WAIT -
tcp 0 0 172.23.99.211:34501 172.23.99.39:8094 TIME_WAIT -
...