DNS: BIND slave won't pick up deletes (only) made via nsupdate on master

I'm a bit perplexed on an issue we're facing. We have a master and slave DNS server running BIND which have been working for many years. We also have a DHCP server running dhcpd and successfully updating (add/delete) ddns via nsupdate.

Recently, we have a need to run nsupdate from another server. Everything works via our nsupdate commands except the following scenario: 'delete' commands (sent not by dhcp server, only) do not propagate to the slave DNS server.

Other scenarios:

• 'delete' commands (sent not by dhcp server) do propagate to the master, and the record file and serial number is updated.

• 'add' commands (sent not by dhcp server) propagate to both master and slave correctly.

• 'delete' commands sent by the dhcpd server (seen in ddns.log) propagate to both master and slave correctly.

• once a 'delete' command (sent not by dhcp server) propagates to the record file on master, 'rndc freeze <zone>' + 'rndc thaw <zone>' pushes it to the slave successfully. (but we shouldn't have to to this -- and don't need to for 'add' commands)

Any ideas on this? Is there a way to see the nsupdate commands our DHCP server is sending, since those are successful? Or any other way to see why this is happening?

Thanks.

EDIT: some clarification on the source of nsupdate command scenarios