Hi laschmoove,
Thanks for your inquiry.
Unfortunately, as you point out, with the current version, it is not possible to
do a frame-based timeout.
As Anthony suggested before, it would need to be implemented as part of your
jobtype backends, but it won't be trivial. You'd need to use signals (SIGALRM)
and/or spawn a monitoring thread/process that will communicate with and
watch your main rendering thread, so that it can kill the main thread when it's
been sitting on a single frame/agenda for more than N minutes.
The other alternative, as Anthony also mentioned, would be to set up your
jobs so that each individual subjob will serve a single frame-- i.e. you'd submit
as many subjobs as there are frames-- and use the subjob timeout.
You can automatically tell your jobs to expand their subjob count to the number
of frames by setting the "expand" and "disable_cpu_limit" flags when you
submit your jobs.
