Inactive instance but no deaths and no logs

hi all, i'm having a troublesome deployment week !

we have an app that is running with 8 instances configured. every now 
and then (at least a couple of times per week) we get the odd instance 
become completely unresponsive. the strange thing is that there is no 
output in the logs, and monitor records no deaths. if i try and stop 
the instance in monitor nothing happens, but i can delete it. however 
if i then add a new instance and try and start it up monitor will 
complain that it's allocated port is still in use.

i'm a little puzzled as to how to debug this one, but what's more 
annoying is that it is virtually impossible to pick up the error as 
there is no indication in the logs that something is wrong, and because 
the instance doesn't die monitor doesn't send me an email! we only 
noticed it because a user reported the site being slow, which was 
actually due to them being allocated the unresponsive instance upon 
their first request and the adaptor waiting 30 seconds before shifting 
them to another instance.

any pointers would be most appreciated.

thanks, simon