We have 14 proxies in the environment. In the "administration" -> "Proxies" tab we have an incorrect number of per proxy items. For readability - we will focus on a single proxy (the wrong number is on all proxies).
In the proxy tab we have:

The number of hosts is correct, but the number of items (1898) is definitely too small. Additionally - "Required vps" seems to be ok.

We started the debug by checking the query "GUI -> server" by tcpdump, as proxy status data is taken directly form servers trapper process. The answer has been shortened to include the most important information:

(...)
"item stats": [
(...)
  {
                "attributes": {
                    "proxyid": 30913,
                    "status": 1
                },
                "count": 17
            },
            {
                "attributes": {
                    "proxyid": 30913,
                    "status": 0,
                    "state": 1
                },
                "count": 57
            },
            {
                "attributes": {
                    "proxyid": 30913,
                    "status": 0,
                    "state": 0
                },
                "count": 1841
            },
      (...)
      ] (...)

Of course, 1841 + 57 = 1898, which agrees with the above number of items (from proxies tab). So we have ruled out a problem on the GUI side.

We then checked the configuration on the proxy side. in short - We noticed the correct value in the logs when increasing the debug level:

(...)
411987:20240314:111800.365 DCsync_configuration() items      : 396984 (1760203 slots)
411987:20240314:111800.365 DCsync_configuration() items_hk   : 396984 (1760203 slots)
411987:20240314:111800.365 DCsync_configuration() numitems   : 338924 (1760203 slots)
(...)

So we assumed correct "server <-> proxy" communication and we have ruled out a problem on the PROXY side.

Returning to the "GUI -> server" communication, the response returns wrong values, so we suspect an error on the zabbix-server side, and in particular - communication with the cache.
After a quick review of the code, we suspect the dc_status_update() function in the dbconfig.c file, where, for example, you can see the counting of the number of items:

(...)
if (NULL != dc_proxy_host)
dc_proxy_host->items_active_normal++;
(...)
if (NULL != dc_proxy_host)
dc_proxy_host->items_active_notsupported++;
(...)

Unfortunately, these are just suspicions and we are not sure where the error is.