Concurrency connections fail

[brenoacp]

Hello everyone,
Yesterday we went through a situation with our Zimbra server where I would like your opinion.
Our Zimbra is virtualized, with 32 GB of RAM, 24 Vproc, release version 8.7.1. GA. 1670. Ubuntu 16.64 UBUNTU16_64 FOSS Edition.
Users have begun to complain that the emails do not leave the Outbox.
We started analyzing we noticed a high use of processing, even a simple connection via Telnet on port 587 took forever to connect.
Analyzing the logs, we noticed that there were many connections from an internal ERP test server sending many emails per minute.
The logo said that the concurrent connections had exceeded the limit and were being ignored, total connections from that server: 51.
Our settings are as follows:

zimbraLmtpNumThreads 30
postfix_lmtp_destination_concurrency=30
zimbraClamAVMaxThreads 30
mailboxd_java_heap_size=6144
mailboxd_java_heap_memory_percent=10

The doubt: why the high processing rate, affecting the other connections?
Is there a way to avoid this situation without affecting the other connections if this case occurs again?

Best regards,
Breno Padovan