That’s frustrating. Are you confident your Core is up to snuff? Swap for another switch(es)? The cheap system on a chip (SoC) switches can be a nightmare.
Roon says it’s demanding on the network but it’s nothing compared to an active enterprise network but then few run enterprise-class gear at home.
Core is I5, 3.2G , 6 core, 8Gb with system on 256Gb SSD/switches are, currently, TPLink (previously Netgear) all unmanaged. Core hardwired to router. Wi-fi is via mesh network with good signal everywhere.
Connection to core, mainly from iPad Pro client, drops frequently then Client struggles to find the core to reconnect even with good signal and careful Wi-fi channel selection.
Which mesh WiFi make?
Linksys Velop Dual, 4 nodes “strategically” placed & operating in bridge mode. Router BThub6 handling DHCP etc…I’ve tried about 6 different routers (some very expensive).
Not sure if this is what the OP is looking for but PingPlotter is a handy tool. Here is a vid (start at 2:20) does a good job explaining it
Thanks for the “heads-up” I’ll give PP a try 
EMCO Ping Monitor is another that is free for 5 hosts. I use it to track my primary DNS forwarder as a gauge of my cable modem node load.
Do you have a local DNS server? I use Pi-hole on a RPi 4b. Solid. Recommended.
I been using NetAnalyser Pro (iOS)…no dropped packets yet 
I may provide local dns once current issues are resolved
Thnx.
Are you pinging your Core?
It’s possible to never drop imcp but multicast could be dropped by a device and cause delays in the client finding the Core.
I’ve forgotten what Roon uses to discover and find within the subnet, MDNS? some kind of multicast I believe.
Here Danny says: We only use mDNS for Airplay discovery.
We also use multicast for our own discovery protocols (not entirely, but we augment with multicast).
Sorry about the delayed response, the day job is getting in the way again.
Pinging the Core is working fine, no dropped packets and averaging 25mSon initial checks but I need to check when the connection to Core drops
25ms is a good long time on a local network. Ok for the first packet if the client needs to ARP but the very next packet should be 2-3ms with the client on wireless. Wired to wired should be sub ms like 0.1ms.
I think that routing:- Client (iPad) to Mesh Node X to Mesh Node Main to Core, introduces delay but I can have a play with that.
Since the removal of the socket fault (first post) the situation has improved drastically, the main irritation now is the iPad dropping connections to the core. Sometimes resuming immediately other times searching for the core for lengthy periods.