4.5Gb Slapcat after server error

[JJ <jj@thecloud.org>]

I've been having some problems while importing some records into my ldap
server and the server crashed with error (23 november)

Nov 23 12:06:55 berlinzoo slapd[29709]: ====> cache_add_entry( 184299 ):
"cn=157041,ou=non iscritti,ou=<CUT> ": already in id cache
Nov 23 12:06:55 berlinzoo slapd[29709]: cache_add_entry_lock failed
Nov 23 12:06:55 berlinzoo slapd[29709]: cache_add_entry_lock failed
Nov 23 12:09:09 berlinzoo slapd[29709]: ldbm: ==> page 3266: illegal
page type or format
Nov 23 12:09:09 berlinzoo slapd[29709]: ldbm: ==> PANIC: Invalid argument

When restarted, sometimes the server crashes so I've read some on google
and decided to db_recover and then restart.

Yesterday (27 November) the openldap server worked "quite" well, but
sometimes still crashes:
Nov 27 08:09:53 berlinzoo slapd[8529]: ldbm: ==> page 3267: illegal page
type or format
Nov 27 08:09:53 berlinzoo slapd[8529]: ldbm: ==> PANIC: Invalid argument
(infact it does not crash, my application just hangs and cannot use ldap
server anymore)

I've noticed that my daily backup (I know, I'm stupid, I should have
restored that backup the next day and not wait till now) since 24 is
getting 2Gb (but just because it lock for max file size reached
otherwise goes over 4Gb).

What I think is that there should be a kind of "circular" reference
because normally the backup is 23Mb, not >4.5GB.

Anyone can suggest a correct procedure to get it stable without
restoring the 23's backup file ?

Thank you very much
Julien