Capture post-hot-loop-fix production PFE dump and profile

Capture a new comparable primary-PFE heap dump and 60-second managed thread/GC/allocation trace from Ithaca after deployed hot-loop PR #2574. Preflight assumption that lifecycle PR #2575 was also deployed was disproven during analysis: image 0792bb8 contains f80f2847e/#2574 but not f25adfe7c, and Gitea #2575 remains open. This capture cannot validate the deadline-queue change. Package dump, trace, metadata/timestamps and checksums in one persistent-host ZIP; download using parallel SSH chunks and verify remote/local SHA-256. No restart, config change or allocator experiment. Preflight 2026-10-05T14:45Z: image 0792bb89c4662faf2448eabaabea8e9e7924583f, PFEs up ~11h; primary container db44f1e3512b uses 16.34 GiB and ~542% CPU at point sample. Host has ~43 GiB available RAM, 7 GiB free swap, 420 GiB disk; PC has 74 GiB free. Verified outcome: local archive /home/mathias/Downloads/pfe-memory-20261005/ithaca-playfield-20261005-memory-capture.zip is 18,172,695,094 bytes and contains heap dump 18,081,497,088 bytes, 60-second nettrace 91,195,826 bytes, metadata/timestamps and inner checksums. Remote zip -T passed. Eight-stream SSH chunk download completed in 668s and local/remote SHA-256 match: ed646e75476fc96cf5ee6490a75db348221e442067f81b5c85220d5d7eefc939. Selected PFE remains running from unchanged start time with RestartCount=0 and OOMKilled=false; final VmRSS 17,110,516 kB / VmSwap 71,340 kB. Temporary in-container dump/trace/tool directory removed. Persistent host backup retained at /home/mathias/pfe-memory-20261005/.