Memory consumption of tcp-disco-msg-worker

Hi

A few months ago we installed the Ignite 2.16 as cluster with 3 server nodes.
All caches that we have configured to use heap memory (property name="onheapCacheEnabled" value="true).
Also we add memory limits: -Xms1G -Xmx1G.

So, our OS system monitoring detected too much memory consumption by ignite - about 1.5Gb
I connect to one node by VisualVM and made the memory sample.
I found that thread with name 'tcp-disco-msg-worker-[some-uuid] allocate too much bytes.

Do you know what is the reason for solid allocation of tcp-disco-msg-worker memory?

Could it be a memory leak?

PS. After restart server node memory consumption coming back to normal

Start with the reported VisualVM sample for tcp-disco-msg-worker and the Ignite 2.16 cluster configuration, including onheapCacheEnabled and the 1G heap limits. Compare memory behavior before and after restarting a server node; done means establishing whether the worker retains memory and documenting a reproducible cause or the evidence needed to investigate further.

Written by the indexing model from the issue text.