[Nagiosplug-help] Inexplicable UNKNOWN state of custom plugin displayed in web interface

[Ralph.Grothe at itdz-berlin.de]

Hello,

I have a nasty bug that I cannot resolve.

I have written a Perl plugin that runs snmp checks against
environmental sensor devices
to poll the temprature at various spots in the data centre.

As far as I can see I have obeyed the prerequisites a plugin must
meet.
It outputs a single line with status and values
as well as the correct return code.

When I execute it from the shell as nagios user it behaves as
expected.
e.g.

$ /usr/lib/nagios/plugins/check_sensor_temp.pl -H sensor_r06 -C
public -w 17:27 -c 10:30
SENSOR OK - temp: T0=21.40 degC T1=24.50 degC 
$ echo $?
0

The plugin's output is also correctly displayed as above in the
web interface's status information field.
However, this is not true for the bg coloring (which is orange)
and 
the text in the table field that displays the status of the last
check, 
where it reads UNKNOWN,
just as if nagios had received an exit code of 3

In the nagios log in the filesystem I can only see appearances of
my nothing changing forced
external commands from reschedule check clicks via the web
interface,
like this (but nothing else):

[1190389354] EXTERNAL COMMAND:
SCHEDULE_FORCED_SVC_CHECK;sensor_r06;SRZ_Temp;1190389352

Trying to look at the event log from within the web interface
stalls my browser
because the log file is obviously too long as I get a lot of
check results for passive checks
from distributed servers.
Maybe a quicker log rotation scheme could alliviate, but I find
event log CGI broken
if it hangs my browser.

Can anyone give me some hint why I get UNKNOWN states displayed,
and how I could try to debug this (maybe some strace on the
nagios scheduler?)

Regards

Ralph