diff --git a/issues/kernel/virt-mask-windows-read-nine-censuses-and-eight-windows-lines-for-one-cpu.md b/issues/kernel/virt-mask-windows-read-nine-censuses-and-eight-windows-lines-for-one-cpu.md
deleted file mode 100644
index db4911a6be6..00000000000
--- a/issues/kernel/virt-mask-windows-read-nine-censuses-and-eight-windows-lines-for-one-cpu.md
+++ /dev/null
@@ -1,41 +0,0 @@
----
-status: open
-kind: defect
-opened: 2026-10-03
----
-
-# `virt_mask_windows` read nine census lines and eight windows lines for one CPU
-
-`virt_mask_windows` (`tests/toyos.rs`) judges the capture `judge_virt_job`
-answers, which ends where `===TEST_END unmap_touch` arrived, with
-`irqcensus::windows`: every `irq: cpu` census line has its CPU's windows line
-beside it.
-
-## Measured
-
-A whole guest suite at `bf28c1e38`, whose `virt_mask_windows`, kernel and
-`tests/common/irqcensus.rs` are `origin/main`'s at `c59e09ed6`:
-`cargo test --test toyos-build`, 12 wide on the 14-core dev host at load
-average 43 to 59, exited 1 with this one red in 66.8 s:
-
-```
-FAIL virt_mask_windows: census lines per cpu {0: 9, 1: 9, 2: 9, 3: 9, 4: 9, 5: 9, 6: 9, 7: 9} and windows lines per cpu {0: 9, 1: 9, 2: 9, 3: 9, 4: 9, 5: 9, 6: 9, 7: 8}: a census went out without its windows, or windows without their census
-  FAIL  virt_mask_windows  (10s)
-```
-
-The whole run before it and the one after it, same tree and same command,
-passed it in 8 s and 7 s. The PL011 capture of the red was not kept: the run's
-`target/red-run-serial` holds the x86-64 guests' 16550 logs only.
-
-Two candidates, neither measured: the capture ends at the marker with cpu7's
-last windows line still behind it, or the kernel printed a census without its
-windows.
-
-## Owner
-
-The windows instrument's author (`issues/kernel/the-kernel-is-small-interrupts-post-and-threads-wait.md`).
-
-## What would close it
-
-The cause, from a capture of a red: either the judge reads a capture that ends
-on a whole report, or the kernel's report is shown to pair its two lines.
diff --git a/tests/common/qemu.rs b/tests/common/qemu.rs
index 0b7917e87d4..d8e54a911cd 100644
--- a/tests/common/qemu.rs
+++ b/tests/common/qemu.rs
@@ -538,6 +538,10 @@ pub fn ceiling_verdict(
 /// Collect console output until `done` reads true of the whole capture, or the
 /// guest stops making progress.
 ///
+/// The capture runs past the line that made `done` true to the end of that
+/// line's drain, which can fall inside anything the guest says in more than
+/// one line.
+///
 /// The shape [`QemuInstance::drain_serial`] cannot have: its caller passes a
 /// number of seconds, and a number of seconds is a claim about the host. Here
 /// the wait ends when the guest goes quiet or wedges, so a guest with a twelfth
diff --git a/tests/toyos.rs b/tests/toyos.rs
index 7ccdc6930f9..35e8f50a82f 100644
--- a/tests/toyos.rs
+++ b/tests/toyos.rs
@@ -1426,7 +1426,8 @@ const VIRT_CPUS: u32 = 8;
 
 /// The windows on AArch64: `tests/virtsmpcase` on [`VIRT_CPUS`] CPUs of a
 /// `mask-windows` kernel, whose every hook checks the state it finds, judged
-/// once the case's job `unmap_touch` has ended.
+/// on the whole console once the case's job `unmap_touch` has ended and the
+/// boot has said its last word.
 fn virt_mask_windows(profile: qemu::Profile) -> Result<(), String> {
     let mut qemu = boot_virt_smp(BootOptions {
         profile,
@@ -1434,7 +1435,10 @@ fn virt_mask_windows(profile: qemu::Profile) -> Result<(), String> {
         kernel_features: toyos_build::build::MASK_WINDOWS_KERNEL,
         ..Default::default()
     });
-    mask_windows(&judge_virt_job(&mut qemu, "unmap_touch", UNMAP_TOUCH_SAID)?, VIRT_CPUS)
+    let mut serial = judge_virt_job(&mut qemu, "unmap_touch", UNMAP_TOUCH_SAID)?;
+    // To the boot's last word, said after every census and its windows: the drain that took the job's end can stop between the two.
+    await_marker(&mut qemu, &mut serial, power::SHUTTING_DOWN, "the boot's last word")?;
+    mask_windows(&serial, VIRT_CPUS)
 }
 
 /// Boot `tests/virtsmpcase` as `options` say.