Mongo driver is having multiple timeouts and one of them is
https://www.mongodb.com/docs/drivers/go/current/fundamentals/connection/

serverSelectionTimeoutMS
integer
30000
Specifies the number of milliseconds to wait to find an available, suitable server to execute an operation.

It has a default value of 30 seconds, so it will always try maximum time allowed by Zabbix agent.

The MongoDB timeout for agent 2 plugin should be configurable or at least smaller than default timeout.

In the source, part where it is creating connection options
https://git.zabbix.com/projects/AP/repos/mongodb/browse/plugin/conn.go#300

We can modify the code to allow the user to manually configure timeout.

Here is diff of MongoDB plugin which allows to lower down timeout:

diff --git a/plugin/conn.go b/plugin/conn.go
index e872a78..7cf65bb 100644
--- a/plugin/conn.go
+++ b/plugin/conn.go
@@ -298,6 +298,7 @@ func (c *ConnManager) createOptions(uri uri.URI, details tlsconfig.Details) (*op
        opt.SetHosts([]string{uri.Addr()})
        opt.SetDirect(true)
        opt.SetConnectTimeout(c.timeout)
+       opt.SetServerSelectionTimeout(c.timeout)
        opt.SetMaxPoolSize(1)

        return opt, nil

Now we can use:

Plugins.MongoDB.Timeout=1

Registering this as a bug report because with existing agent configuration, it's not possible to workaround the issue. The poller which is fetching "mongodb.ping" metric will get stuck.