rca-236 INSTRUMENTATION — preserved 2026-08-29 immediately before the post-v0.65.0 reap.
These patches are the ONLY surviving copy. Both source trees are gone.

BASE SHA for both: 0c17eb09d865db1669a28ecfc7907f44f2cb06eb (the golden/conduit-236-r2 head).

  rca-236-win-instrumentation.patch
    4366 bytes  sha256 816692d2561d48acbb58dcb2227ef95cd0f8d396a1882e4e3fcd22ed53898e3b
    4 src files: spt-daemon/src/iobus.rs, spt/src/api/delivery.rs, spt/src/cli.rs,
                 spt/src/roster.rs
    VERIFIED: applies CLEAN at main 4d6007ac (git apply --check, empty output).
    Taken from C:\...\spt-core\.worktrees\rca-236-win before `git worktree remove --force`.

  rca-236-kitsubito-instrumentation.patch
    39106 bytes  sha256 b98580e561a8a89c3377861ec052184811657db7119e1ac2f49d38d887d5350b
    The same 4 src files PLUS crates/spt/tests/io_events_undriven_kinds_e2e.rs
    (830 changed lines, 478 insertions / 402 deletions — the substantial half).
    MEASURED, not assumed: it does NOT apply straight at main 4d6007ac. `git apply --check`
    fails on the TEST file only (that file moved on main after 0c17eb09); the four src files
    apply cleanly under `--3way`. To restore, check out 0c17eb09 and apply there, or use
    `git apply --3way` and merge the test file by hand.
    sha256 verified byte-identical on both sides of the scp from kitsubito.

WHY THERE ARE NO FULL-FILE COPIES. I tried to also preserve the five files verbatim (a patch
that does not apply is weak preservation for an 830-line change). By the time that command ran
— roughly one minute after the diff was taken — BOTH kitsubito trees had already been reaped by
another agent acting on the same pools-released call. The diff had landed; the file copies had
not. The patch above is therefore the whole record, and the near-miss is the reason this file
says so out loud instead of leaving a reader to assume a fuller copy exists somewhere.