Clearing State Table

I've read previous posts regarding the state table not clearing in 
4.1.15.  I have a similar problem under 4.1.28 (as well as .26 and .3).  
I'm using this on Sparc Solaris 10.  Mine's a little different in that 
it does clear when it hits 50% bucket size.  Has anybody ever seen this 
before?  Am I missing something?

In this example I'm hitting the server with one http request at a time 
from one host only.  At most there should only be two active 
connections, 1 http connection and 1 telnet session.

dr2 ~/ipf> ipfstat -s
IP states added:
        16048 TCP
        46 UDP
        0 ICMP
        163521 hits
        16095 misses
        3 maximum
        0 no memory
        0 max bucket
        3 maximum
        0 no memory
        2852 bkts in use
        3924 active
        0 expired
        12036 closed
State logging enabled

State table bucket statistics:
        2852 in use
        49.71% bucket usage
        0 minimal length
        5 maximal length
        1.376 average length

dr2 ~/ipf> ipfstat -s
IP states added:
        16149 TCP
        46 UDP
        0 ICMP
        164537 hits
        16197 misses
        4 maximum
        0 no memory
        0 max bucket
        4 maximum
        0 no memory
        13 bkts in use
        13 active
        0 expired
        16048 closed
State logging enabled

State table bucket statistics:
        13 in use
        0.23% bucket usage
        0 minimal length
        1 maximal length
        1.000 average length