Steps to reproduce:

1. Add 3 hosts
2. Configure an SNMP item with the following parameters:

    * Name: Manager: get metrics
    * Type: SNMP agent
    * Key: manager.get
    * Type of information: Text
    * SNMP OID: walk[enterprises.43925.73.1.1.1]
    * Update interval: 30s
    * History: Do not store
    * Preprocessing:
       Name: SNMP walk to JSON
       Field name: value
       OID prefix: .1.3.6.1.4.1.43925.73.1.1.1
       Format: Unchanged
    

1. Configure a "Dependent item" that depends on "manager.get":

    * Name: ChangesRequests
    * Type: Dependent item
    * Key: enterprises.43925.73.1.1.1.8.1
    * Type of information: Numeric (unsigned)
    * History: 14d
    * Trends: 180d
    * Preprocessing:
       Name: JSONPath
       Parameters: $[?(@.['{#SNMPINDEX}'] == '8.1')].value.first()
    

1. Include both items in a template and link this temlate to 2 hosts
2. On the third host configure a "Calculated item" with the following parameters:

    * Name: ChangesRequests (aggregated)
    * Type: Calculated
    * Key: ChangesRequests_aggr
    * Type of information: Numeric (unsigned)
    * Formula: sum(last_foreach(/*/enterprises.43925.73.1.1.1.8.1?[group="{$MANAGER_HOSTGROUP}"]))
    * Update interval: 30s
    * History: 14d
    * Trends: 180d
    

Set the value of macro {$MANAGER_HOSTGROUP} to the name of the group, that includes only the two hosts.

Result:

See screenshot

Starting from 2025-07-28 16:02:45, the value of the calculated item (first column) does not change, even though the dependent item value on the first host is updating.

Expected:
The calculated item should reflect the latest values from the dependent items.

Possible issue:
The last_foreach(), sum(), or value cache may not be working correctly. I can't find problem itemids in the output of "zabbix_server -R diaginfo=valuecache".