chore(quest): plan #407's follow-ups: a moq checkout per pin, and Python's wake and call cost

quest/m1/ffi-export/README.md

@@ -60,3 +60,4 @@ than `ctypes`; the others call C natively. moq's packages switch to them in moq'
 - [Python bindings](/quest/m1/ffi-export/python.md) - a generated CPython extension, skipping the C ABI as Kotlin's JNI does
 - [Go bindings](/quest/m1/ffi-export/go.md) - Go over the C ABI through cgo
 - [Dart bindings](/quest/m1/ffi-export/dart.md) - Dart over the C ABI through `dart:ffi`
+- [A moq checkout per pin](/quest/m1/ffi-export/moq-checkouts.md) - `just ts` and `just ffi` check out their own moq revisions, so both run at once

quest/m1/ffi-export/moq-checkouts.md

@@ -0,0 +1,17 @@
+# [S] A moq checkout per pin
+
+## Goal
+
+`just ts` and `just ffi` each check out the moq revision they pin in a directory of its own,
+so running both at once in one worktree, such as `just ci` beside `just ffi bench`, builds
+each against its own revision.
+
+## Plan
+
+Found by [#407](https://github.com/moq-dev/rust2/pull/407): `just ffi standin` reads
+`ts/target/moq`, which `just ts moq-pin` checks out at ts's revision, and the two pins have
+differed since [#397](https://github.com/moq-dev/rust2/pull/397). Running both at once fails
+the export with a missing crate. Key the checkout by its revision, or give ffi its own,
+sharing the clone's objects rather than fetching twice.
+
+Public API: none.

quest/m3/README.md

@@ -40,3 +40,5 @@ generic std gap is planned here even when moq never calls it.
 - [An Option of numbers in the result slot](/quest/m3/ts-option-slot.md) - `varint_opt`'s `Option` comes back through the slot with no object per call
 - [Dropping a crate's generic type inside another](/quest/m3/ts-nested-generic-drop.md) - an `LruCache` holding a `LinkedHashMap` drops as rustc's glue does
 - [Fresh writes through a zip, a copy, and a repeat](/quest/m3/ts-fresh-shapes.md) - zipping a `Vec<[T; N]>`, reading a `T` out of `Zip::next`, and `copied` or `vec![x; n]` over a `Copy` type translate
+- [Python's wake on the loop's own thread](/quest/m3/ffi-python-wake.md) - a wake on the event loop's thread resumes the future with no socket pair
+- [Python's call cost](/quest/m3/ffi-python-calls.md) - a call on an object costs what BoltFFI's does, or the gap is explained

quest/m3/ffi-python-calls.md

@@ -0,0 +1,15 @@
+# [M] Python's call cost
+
+## Goal
+
+A call on an object through the exporter's [Python](/ffi/README.md) costs what
+[BoltFFI's](/bench/ffi/README.md#boltffi-comparison) does, 63 ns, or `just ffi bench`
+explains the gap.
+
+## Plan
+
+Found by [#407](https://github.com/moq-dev/rust2/pull/407), which measured about 125 ns on a
+loaded host. Measure on a quiet one first, then profile where the extension spends the rest:
+argument parsing, the handle's lookup, and the contract or GIL work per call.
+
+Public API: none.

quest/m3/ffi-python-wake.md

@@ -0,0 +1,15 @@
+# [S] Python's wake on the loop's own thread
+
+## Goal
+
+A future the exporter's [Python](/ffi/README.md) awaits resumes directly when Rust wakes it
+on the event loop's own thread, skipping the socket pair, and `just ffi bench` measures the
+woken read against [BoltFFI's](/bench/ffi/README.md#boltffi-comparison) 10 µs.
+
+## Plan
+
+Found by [#407](https://github.com/moq-dev/rust2/pull/407): a woken read costs about 21 µs,
+about 5 of them the socket pair's write and read. A wake from another thread still goes
+through it, and the wake-latch and cancel tests keep passing.
+
+Public API: none.