In September 2004, Linux kernel developer Thomas Habets proposed an 'oom_pardon' patch to protect critical processes from the OOM killer. His sysctl would allow administrators to specify processes that should never be terminated during out-of-memory situations, even if it meant triggering a system panic instead.

Andries Brouwer responded with a provocative analogy about an aircraft company that solved fuel shortages by ejecting passengers. This parable illustrated how OOM mechanisms could become problematic - once implemented, they might activate inappropriately or create perverse incentives around which processes get killed.

The mailing list discussion highlighted fundamental tensions in memory management policies. While protecting essential processes like xlock sounded reasonable on the surface, the broader implications raised uncomfortable questions about system reliability and whether introducing kill switches was wise.

This early 2000s debate prefigured modern concerns about automatic resource management in containerized environments, where protecting critical infrastructure processes while maintaining system stability remains an ongoing challenge for kernel developers.