We have some python post-job tasks, and today a bad library import slipped in (missing dll for a dependency).
Sadly, it took out our entire farm due to the python sessions not being sandboxed. Every slave that picked it up threw a c++ runtime error in the slave application itself:
Faulting application name: deadlineslave.exe, version: 6.2.1.49, time stamp: 0x54380381
Faulting module name: ntdll.dll, version: 6.1.7601.18247, time stamp: 0x521eaf24
Exception code: 0xc0000374
Fault offset: 0x00000000000c4102
Faulting process id: 0x15cc
Faulting application start time: 0x01d04a35a64ee6a5
Faulting application path: C:\Program Files\Thinkbox\Deadline6\bin\deadlineslave.exe
Faulting module path: C:\Windows\SYSTEM32\ntdll.dll
Report Id: 2c310e70-b632-11e4-93d9-0cc47a0a744b
Needless to say, python scripting errors being able to take down the entire farm (instead of just erroring the affected tasks) is a major problem