I'm pretty new to Juniper, so maybe the status at my new employer is rare, but it seems a large portion of our switches have errors in their /var partitions, as reported by fsck. Once I noticed the issue (after being bitten by it), I checked through the switches of a 10-floor dorm, with either a single ex4200 or a small stack on each floor, and found that 7 of 10 stacks showed fsck errors on /var on either the routing-engine or backup controller, and one floor had errors on both the RE and BK members.
These errors make the reboots after installing a software update less likely to go smoothly, and worse, seem to need a manual console connection to repair. That doesn't scale well for trying to upgrade several buildings worth of switches.
I'm hoping there's some option to make the device either run an agressive fsck at reboot to fix this; or, (better, I think!) nuke-and-rebuild-from scratch the /var directory as part of the firmware update installations. This would be a painless option during software updates, as since the switch is already offline and rebooting several times in any case.
Ideally, I'd like to be able to load a software update to switches during the workday, and then after they get through to the "reboot required" stage, schedule a reboot for 4am, perhaps with a flag telling it to nuke /var during the software install. (It seems errors in the /var/tmp/ partition are much more rare, so perhaps the system can keep a copy of the new software image here until it has installed it and fixed up /var)