Milter-greylist

I'm testing milter-greylist with four Sun Netra T1AC200, 1GB RAM,
running Solaris 8 (with latest rec. patches added), working as 
incoming MX for a company.

We're using sendmail 8.12.10 and milter-greylist 2.0rc3.

The maillog tells me the one machine received 125000 mails the
last 16 hours, resulting in 2-3 mails per second.
Since activating greylisting the machines load is around 3-4
during working hours.
greylist.db has currently about 500000 lines.

I'm observing some problems which I'm not shure who'm I should
blame for:

Together with the 125000 mails given above I get about 14000
times log messages like this in sendmail log:

Jul  6 00:04:29 xxx sendmail[21218]: [ID 801593 mail.error] j65M4Ig5021218: Milter (greylist): timeout before data read
Jul  6 00:04:29 xxx sendmail[21218]: [ID 801593 mail.info] j65M4Ig5021218: Milter (greylist): to error state

That sounds like milter-greylist isn't answering fast enough
so sendmail is giving up and skipping it. Is this timeout
tunable? What's the default?

Apart from the first problem people get noticable amounts of mail 
with no X-greylisting header at all and sendmail log gives no 
indication that it even tried to talk to milter-greylist, no timeout,
nothing. What might be going on there?


milter-greylist populates the db-directory whith halfway-finished
db dumps, like this:

-rw-------   1 smmsp    smmsp    51392876 Jul  6 16:22 greylist.db
-rw-------   1 smmsp    smmsp    45981696 Jul  6 13:12 greylist.db-XXP6aqei
-rw-------   1 smmsp    smmsp    45793280 Jul  6 12:58 greylist.db-XXk_ayPU
-rw-------   1 smmsp    smmsp    49692672 Jul  6 12:32 greylist.db-XXT1aWDI
-rw-------   1 smmsp    smmsp    47104000 Jul  6 12:18 greylist.db-XXVOaWEx
-rw-------   1 smmsp    smmsp    31637504 Jul  6 12:04 greylist.db-XXMRayKo
-rw-------   1 smmsp    smmsp    46284800 Jul  6 11:50 greylist.db-XXfpaWRf
-rw-------   1 smmsp    smmsp    46071808 Jul  6 11:36 greylist.db-XXFSaqX0

It looks like the dumping process is interrupted sometimes without
beeing able to resume and finish properly. 
What could be done against this behaviour, apart from setting up a
cronjob for removing the incomplete db-files?  :-)

The four machines have constant trouble with syncing the databases.
I get queue overflows every few minutes.

Are all of the problems caused by (too) high system load?
The machines are doing nothing else but mail routing and activating
milter-greylist more than doubled system load. Is this normal behaviour?

Are there known problems with sendmail's milter interface, 
are there improvements in this area with sendmail 8.13?

Any helpfull ideas? 


Thank you!

nils-henner