The interesting part is:
/usr/bin/snmpbulkwalk -t '5' -r '2' -v2c -c '<private>' -Pu -OQUs -m EtherLike-MIB -M /opt/observium/mibs/rfc:/opt/observium/mibs/net-snmp 'udp':'<hostname>':'161' dot3StatsEntry
[...]
dot3StatsDuplexStatus.508 = fullDuplex
dot3StatsDuplexStatus.509 = fullDuplex
dot3StatsDuplexStatus.510 = fullDuplex
dot3StatsDuplexStatus.511 = fullDuplex
dot3StatsDuplexStatus.513 = fullDuplex
dot3StatsDuplexStatus.514 = fullDuplex
dot3StatsDuplexStatus.515 = fullDuplex
dot3StatsDuplexStatus.516 = fullDuplex
dot3StatsDuplexStatus.517 = fullDuplex
dot3StatsDuplexStatus.519 = fullDuplex
dot3StatsDuplexStatus.523 = fullDuplex
dot3StatsDuplexStatus.526 = fullDuplex
dot3StatsDuplexStatus.528 = fullDuplex
dot3StatsDuplexStatus.529 = fullDuplex
[..]
Do you see any possibility to circumvent the bug?
(Meanwhile I will open a case @Juniper, but I assume they are as big as a giant tanker. Mostly you need a speedboat for making customers happy)
Access to a device can be provided.
Indeed, this is 100% a JunOS bug, typical. There's no way of us knowing this data is wrong, so you'll have to exclude the devices which exhibit this from the alert definition.
Probably easiest done with "hostname != <hostname>" lines, i guess.