As a reference point, we currently have 2,704 edges in our VMWare/Velocloud deployment.  With the 8 out of box item prototypes, that's 21,632 dependent items just for edges.  2 links per edge (a primary and backup connection) is our standard, which means 5,408 links which gives us another 54,080 dependent items with the out of box VMware SD-WAN VeloCloud by HTTP template.  We disabled them trying to get us under the limit but Peers and path metric item counts basically scale with edges so you can see where this is going.

While the proposed item limit increase in 7.2 should at least allow all items to be created, it's still a bit of a UI/Navigation nightmare.  The otherwise functioning normally web server config had to be changed (upgrading memory limits, etc) to handle viewing items, latest data, and so how just for this host.  Trying to find link data or graphs for a specific edge is painful.

It's also important to note that unless you're aware of the item limit, you likely not aware there's an issue as there's nothing in the UI telling you that you're missing data.  With 30,000 items there's no practical way to audit if anything is missing.  You can see you've got links, edges, gateways, so you assume all is well.  Last but not least, the script struggles to complete even with timeouts increased to max as it's returning 300MB+ of data.