Capture post-navigation-fix production PFE memory snapshot

User reports production PFE memory growth continues after navigation ownership fix PR #2555. Capture one new heap dump on Ithaca and copy it to Mathias' PC for offline comparison with the verified 2026-10-03 capture (tsk_wn8ck8pxzj / investigation tsk_73fyezjdwp). No live allocator experiments, configuration changes or restarts. Confirm collection timing/player warning before pausing a PFE. Prefer the primary PFE for role consistency with the previous capture; retain a persistent host backup outside Docker. Use standalone downloaded diagnostics if shipped tools cannot run; transfer with parallel SSH chunks and verify SHA-256. Verified outcome: local ZIP /home/mathias/Downloads/pfe-memory-20261004/ithaca-playfield-20261004-memory-capture.zip (18,680,043,672 bytes) contains heap dump 18,586,329,088 bytes, 60-second nettrace 93,712,354 bytes, timestamps, metadata and inner file checksums. Profile combines dotnet-sampled-thread-time and gc-verbose. Archive created with ZIP stored mode, passed remote zip -T, downloaded using eight parallel SSH chunk reads, reassembled and matched remote/local SHA-256 435e3c0c1b12538bebd3304a83d6b92f54514d464c1d4521837d7351173c400d. Persistent remote backup retained at /home/mathias/pfe-memory-20261004/. Selected PFE still running from the same start time with RestartCount=0 and OOMKilled=false; final RSS ~18.0 GiB / swap ~1.6 MiB. Temporary in-container dump, trace and tool files removed. Preflight 2026-10-04T19:26Z: image tag 16e2352911449e4ee33d86addcf6a53f42075eff, PFEs up ~6 hours using ~15–16 GiB each; host ~41 GiB available RAM and 455 GiB disk free; PC ~80 GiB free. Selected primary container a4eaa0498f6b. Snapshot collection not started; awaiting timing confirmation.