[c-nsp] Spurious Memory Access and DHCP

[Osama I Dosary]

We want to use DHCP over VPDN (running IOS12.3-9c on a 7301.) We need it
to set DNS domain-suffix on clients.
We've encountered large spurious memory access when we started using
DHCP for this purpose. When I increased the address range in the pool
and then did a "show run" it would cause the session to hang, and
sometimes cause this error message:
        % The DHCP database could not be locked. Please retry the
command later.
And sometimes "write mem" would take a minutes or so. Btw, CPU util is
very low ~2%.

This lasted for over 15 minutes, and we started loosing sessions until
all were gone. I couldn't find info about this behavior, so I reloaded
into 12.3-5a.B4, thinking it might be an undocumented bug in 12.3-9c,
but we still experienced high spurious memory access (~600 within 4
hours) afterwards.
When we switched back to using local-pool instead of dhcp-pool, things
get back to normal, with no spurious mem access.

I found nothing when I googled for this error message. Pasting "show
align" to the output interpreter gave me that it was a software bug, but
not much more information. I couldn't find any matching bug using bug
toolkit.

Could DHCP and spurious mem access be directly related -- Meaning that
could DHCP be malfunctioning doing the spurious access, or could this be
some irrelevant hidden bug that was triggered by DHCP and could be
triggered by some other feature?

Has anyone encounter anything like this, or have ideas about what this
could be? And/or fix DHCP?

TIA
-Osama