hostname problem with snmpd

[Dave Mason]

Hi,
I'm having a strange problem where a user is changing the hostname to 
something that doesnt resolve, and the agent does not respond even after 
the hostname is restored to its proper value.  The problem can only be 
fixed by deleting the persistence data files and restarting.  Here's the 
full sequence:

1. set hostname to something that DNS can't resolve, ie
   hostname xxx
2. start snmpd as usual.  The same problem also happens if the agent is 
already running before the hostname is changed.
3. try some SNMP GETs.  They fail as expected.
4. restore hostname to it's proper value, which DNS can resolve
5. restart snmpd with persistence data files from before.
6. try SNMP GETs.  agent still does not respond.
7. to fix. stop the agent, delete persistence data files, and restart.

The obvious answer is to avoid screwing up the host name, but customers 
being customers, they want a solution.  They believe the agent should 
work fine once the hostname is restored, and they don't want to lose 
their persistence data.  Is it possible that the hostname is written 
into the persistence files somewhere, causing problems once it's changed?

We have had some other issues with persistence files, so I'm also 
curious as to what is going on.  We're running on Red Hat 7.3, agent++ 
version 3.5.3a, snmp++ version 3.1.6c.

Regards,
Dave