We are currently building a core dump based scheme to monitor crashes on our deployed devices. Devices use a different architecture than our development environments, and the deployed devices are often accessible only with limited bandwidth, so a live shell session can be more painful than waiting for the core dump to come back.