Firstly - this is not a SolarWinds problem. I am just wondering if anyone else has seen or heard of similar issues.
We finally got around to upgrading to the latest and greatest versions on our Orion platform and add two additional poller engines in to take the strain off.
It looks great and the new features have gone down well with our Support teams. 
That aside - we first started noticing problems whilst building up the APEs, when RDP session kept being dropped, I put it down to patching (which we then did).
The new version of the Desktop client has it's own alerting for connection issues to the Orion server and these are popping up and down on the support teams desktops, at a rather annoying frequency (along with the response team's klaxon noise going off).
We have also been receiving an increased number of false packet loss alerts, for a variety of devices (virtual, physical, WMI, agent and SNMP) and from all over the environment, appearing at the same time and then clearing down at the same time.
If this was for some of the customers' sites we manage, on WAN links or some remote offices that occasionally power down for the weekend, packet loss is occasionally seen, but on our core management VLAN? This is not right.
So I raised a ticket to internal teams to investigate, we dug through event logs and log files, with common errors of disconnections and other network related issues.
Testing from networks came back negative (as we all know it's never the network! 
), however just a simple ping from my laptop to my Orion server is showing obvious packet loss.
Then one of our platform chaps found an article on the VMWare knowledge base about an known issue with packet loss on heavily utilised VMXNET3 vNICs - https://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=2039495 , which appears to be likely to be related. I have to say this is one of those wonderful needle in a haystack problems, that has been passed around the support teams trying to find the cause.
So if this is the issue, it appears that when busy these vNICs, start to drop packets under heavy loads. It does however state that this may related to filtering, which I don't believe we do on our Orion (we no longer use FoE).
Unless I am forgetting something, but sure Orion doesn't filter the packets (correct me if I am wrong please). So I will assume that it is just load.
(Why it is more apparent after I have added two APEs and reduced polling by 1/3, is also odd - surely the vNICs are quieter now?)
The problem we have is that we cannot recreate this issue in test/dev, as we can't generate a suitable load to recreate the issue, so we are just about to test in production.
So, has anyone else out there ever come across this problem, or even gone through the steps of increasing the buffers?
Did it work?
Can anyone else add anything, that might help us? 
Oh, and our Orion platform is multi tenanted with paying customers, so the thought of testing this in production fills me with dread...
Thanks,
Rich
