A bug in 2.0REL? - WAS: falling hit ratios?

I wrote about two weeks ago about a falling hit ratio problem. I think I've
nailed down the problem.

If you reduce the size of a /cache dir, like from 3.3g to 3.1g then run
squid -k reconfigure it falls apart, with the hit ratios steadily falling
Any idea why? I thought it was supposed to automatically purge old data etc?

System:
P2-450/256M/RedHat 5.1/2.0RELEASE/IPFW/ Cisco 2514 route-map

Log:
908807475.161 52 204.203.237.144 TCP_MISS/302 165 GET
http://www.cbvcp.com/ - DIRECT/www.cbvcp.com -
908807475.756 418 204.203.237.144 TCP_MISS/200 24884 GET
http://metro.cbvcp.com/cbvcp/ - DIRECT/metro.cbvcp.com text/html
908807476.644 70 204.203.237.144 TCP_MISS/304 119 GET
http://metro.cbvcp.com/cbvcp/graphics/lighoff.gif - DIRECT/metro.cbvcp.com -
908807476.715 32 204.203.237.144 TCP_MISS/304 119 GET
http://metro.cbvcp.com/cbvcp/graphics/newsico.gif - DIRECT/metro.cbvcp.com -
908807477.265 143 204.203.237.144 TCP_MISS/304 119 GET
http://metro.cbvcp.com/cbvcp/graphics/entertainico.gif -
DIRECT/metro.cbvcp.com

908807489.901 33 204.203.237.144 TCP_MISS/302 165 GET
http://www.cbvcp.com/ - DIRECT/www.cbvcp.com -
908807490.308 298 204.203.237.144 TCP_MISS/200 24884 GET
http://metro.cbvcp.com/cbvcp/ - DIRECT/metro.cbvcp.com text/html
908807491.076 117 204.203.237.144 TCP_MISS/304 119 GET
http://metro.cbvcp.com/cbvcp/graphics/lighoff.gif - DIRECT/metro.cbvcp.com -
908807491.772 51 204.203.237.144 TCP_MISS/304 119 GET
http://metro.cbvcp.com/cbvcp/graphics/newsico.gif - DIRECT/metro.cbvcp.com -
908807492.014 52 204.203.237.144 TCP_MISS/304 119 GET
http://metro.cbvcp.com/cbvcp/graphics/entertainico.gif -
DIRECT/metro.cbvcp.com
Received on Tue Nov 03 1998 - 11:50:26 MST