diff --git a/harness/mame/README.md b/harness/mame/README.md
index 49797f3..90c21ad 100644
--- a/harness/mame/README.md
+++ b/harness/mame/README.md
@@ -22,10 +22,15 @@ are written down rather than assumed.
    knowing why this keeps happening: `scn2681.cpp` **does not exist** (the SCN2681
    lives in `mc68681.*`), so a search for it walks the entire disk before giving up.
 
-2. **Stage only your own paths.** `git add <path> [<path>…]` — never `git add -A`,
-   never `git commit -a`. Several workers share this repository at once, and a `-A`
-   sweeps someone else's in-flight files into your commit. That happened twice in one
-   day. If it happens, unstage and `git commit --amend`, and say so.
+2. **Stage only your own paths — and do not edit a file someone else is holding.**
+   `git add <path> [<path>…]` — never `git add -A`, never `git commit -a`. Several
+   people and workers share this repository at once, and a `-A` sweeps someone else's
+   in-flight files into your commit. That happened twice in one day, and then a third
+   time because the *orchestrator* edited this very README while a worker held it — so
+   note that an **uncommitted** edit of yours can be swept into someone else's commit
+   just as easily as an unstaged file. If you are about to touch a file another worker
+   owns, don't: say what you want changed instead. If a collision does happen, unstage,
+   `git commit --amend`, and say so.
 
 3. **One sample is not evidence.** "It never paints", "the driver only models 256 KiB
    of RAM", and "it is stuck" were all *wrong*, and all three came from a single
@@ -42,8 +47,9 @@ are written down rather than assumed.
 5. **Do not fake it.** No responder that feeds the firmware the answer a test wants.
    Where a real device is unknown, model what the documentation actually supports and
    **report the gap**; where a stand-in is genuinely unavoidable, make it **opt-in and
-   name it for what it is** — see `VT1200_IO_PATH_STUB` and `VT1200_SKIP_LEVEL1`, both
-   of which are documented divergences, not quiet assumptions. And the **ROM stays
+   name it for what it is** — see the **divergences table** above: `VT1200_IO_PATH_STUB`,
+   `VT1200_SKIP_LEVEL1`, and the tablet default are all documented divergences, not
+   quiet assumptions. And the **ROM stays
    byte-identical** (`md5 b46a3e442ccb86134d2356bd70d22fd2`): patch at runtime if you
    must, never the file.
 
diff --git a/harness/mame/run.sh b/harness/mame/run.sh
index 2bc936f..41c84dc 100755
--- a/harness/mame/run.sh
+++ b/harness/mame/run.sh
@@ -62,9 +62,9 @@ if [ $# -gt 0 ]; then shift; fi
 usage() { awk 'NR>1 && /^set -euo pipefail/{exit} NR>1{print}' "$0"; }
 
 # common: correct rompath, no sound (the driver has no audio), no licence screen.
-# The lightgun options give MAME an absolute host pointer for the VSXXX-AB tablet
-# (VT1200_TABLET=1); they are harmless when the tablet is not selected, and they do
-# NOT grab the pointer.
+# The lightgun options give MAME an absolute host pointer, which the DEFAULT VSXXX-AB
+# tablet needs and the VSXXX-AA mouse (VT1200_MOUSE=1) also uses to follow the host.
+# They are harmless either way, and they do NOT grab the pointer.
 COMMON=(-rompath "$ROMPATH" -sound none -skip_gameinfo -lightgun -lightgun_device mouse)
 
 case "$MODE" in