todlando -- WINDOW START, on doyle WORD 10:44:51Z (his census 0, post-merge run 34589408286 GREEN on all five jobs). This message is my FIRST act of the window; nothing has been launched yet.

SUBJECT: 3e7eaf67, lane feat/304-remote-friction-product, worktree .worktrees/304-product, pool .worktrees/304-product/target, label todlando-304-product. Tree dirty only at docs/adr/0058-attachments-are-pull-model.md (not compiled).

ADMISSION, resampled at the word (10:45Z), my own reader, the same instrument hertz and I agreed on:
  total processes 560; BOTH positive controls TRUE (one process wearing both keyed properties -- copied to the name cargo.exe AND launched from my pool's debug\deps -- alive across the poll, killed after, residue verified ABSENT).
  path-keyed hits in MY pool deps, excluding the control: ZERO. Nothing foreign is executing in my pool.
  name-keyed hits, excluding the control: 26640 and 46072 rust-analyzer.exe (the IR-100 caveat -- the analyzer resolves my worktree manifest and takes my build lock; it is not a peer's builder), 39260 and 43280 rustup.exe (shims, no pool path; consistent with hertz building in HIS pool during my window, which doyle authorised).
  blind 3, by identity, the same three as every prior census on this box: pid 0 System Idle Process AccessDenied, pid 4 System EmptyPath, pid 236 EmptyPath. Counted as blind, not skipped.
  FREE SPACE 164.3 GiB against a 96 GiB admission floor; my pool is 18.8 GB now. Stop conditions: <=32 GiB free or >=64 GiB growth.

PLAN, two arms, one attempt each, no retry, 600 s cap TOTAL, summed budget ~180 s expected (arm 0 build ~40 s + one cell; arm 1 build ~25 s + full --lib):
  ARM 0, the negative control FIRST: delete the slash-pair conjunct from docshost::relabel_location, build, run ONLY docshost::tests::an_absolute_redirect_location_passes_through_untouched. PREDICTED RED at the protocol-relative assert with left /rig-b/hfenduleam/docs/ against right //elsewhere.example/hfenduleam/docs/. Arm 0 GREEN refutes the CELL, not the product, and I would say so.
  RESTORE the conjunct and verify the tree is clean against 3e7eaf67 BEFORE arm 1 -- a rig step, not an afterthought.
  ARM 1: cargo nextest run -p spt-daemon --lib --success-output immediate. Predictions as amended and already accepted: exactly TWO reds, neither the rider -- adapter_facet_and_disambiguated_alias_share_one_contained_reference PANIC at webserve.rs:1133, index_escapes_html_encodes_urls_and_reports_real_source_paths ASSERT FAIL at :1197 (not a panic; :1203 and :1207 masked behind it). Green: my four docshost cells, webproxy::tests::under_node_prefix_replaces_only_the_label, and everything else green at 0845e68d. P5 holds by oid 798ef40c, webserve.rs byte-identical at 0845e68d and 3e7eaf67.
  --success-output immediate is on because two of my predictions are about WHICH LINE fails, and a nextest green prints a passing test's output nowhere.

END will name run/passed, every matched case name, and each prediction line against its result -- including anything my own filing got wrong.