Regressions used to test NOV caching:

All the following with OVER configured on the NNTP server (where used)
and a system wide local .overview type file

caching = OFF, read local spool        => should use system wide nov file
                                          should not read or write private cache

caching = OFF, read via NNTP           => should read via OVER only
                                          should not read or write private cache

caching = ON, read local spool         => should use system wide nov file
                                          should not read or write private cache
caching = ON, read via NNTP
        private cache not present      => OVER info used to create private cache
        private cache is present       => private cache read and used
        partial private cache(*)       => private cache read and used, completed
                                          using OVER and written back

The following with local .overview nov files NOT present...
caching = OFF, read local spool        => no cache read or written

caching = ON, read local spool
        private cache not present      => private cache created as best as
                                          possible from headers
        private cache is present       => private cache read and used
        partial private cache          => private cache updated with missing
                                          header info

(*) partial = end missing - There is certainly no way at present for 'holes'
    to be filled