I'm probably not the best owner for this since I don't own recovery images, and factory shims have its own way of debuggibg.

I was thinking of having a common default for all 

what is the factory flow now? drop to a console and get stuff by hand?

Currently recovery will put a log of shell debug logs (+x), which is good enough for most of the time... or not?

For factory images, we... 
 (1) add cros_debug to spawn an extra shell for debugging
 (2) if WP=off, scripting (installation) execution error will also spawn a shell to debug

And no, usually we don't need to look at coredump to figure out what goes wrong. The script execution log is enough for factory.

i see recovery has save_log_files, but so does factory_shim.  but if you don't care about that, then let's ignore it for now.

how about we:
- in common/init.sh:
  - add an init_coredumps to set an rlimit to something reasonable (let's say 100MB) and have coredumps save to /log/crashes
  - always call that from `initialize`
- in recovery/init, change save_log_files to copy over the coredumps last

do you see a problem with that ?

I'm ok with adding rlimit, but can I leave it to someone else since my plate is full for all factory stuff :)