Remove what did not earn its place: kernel patch 0005, the getty drop-in, four probes

.github/workflows/lab.yml

@@ -36,10 +36,7 @@ on:
           - boot:userspace
           - boot:systemd
           - boot:waits
-          - boot:jobs
           - boot:ssh-first
-          - boot:exits
-          - boot:mmio
           - boot:ssh-load
           - boot:page-cache
           - boot:firmware

boot/Taskfile.yml

@@ -193,20 +193,6 @@ tasks:
     cmds:
       - SPIN_BOOT_WAITS=1 go test ./boot/ -run '^TestBootWaits$' -count=1 -v -timeout 20m
 
-  jobs:
-    desc: >-
-      PID 1's job queue as it changed during the boot, recorded live over systemd's private
-      socket by boot/jobwatch, a diagnostic binary built into each boot's overlay: what was
-      waiting and what was running every BUCKET ms, the boot:waits gaps with the jobs running
-      through them, jobs that did not end done, and the jobs the first SSH login after the boot
-      sets off. Interleaved with boots
-      without it, and the first table is what recording cost. Needs KVM, a built release and
-      sudo. REPS= boots of each, CPUS= vCPUs of both (3 gives the recorder room of its own),
-      LINGER= how long it records after startup, RAW=<dir> keeps the dumps.
-    deps: [':tools']
-    cmds:
-      - SPIN_BOOT_JOBS=1 go test ./boot/ -run '^TestBootJobs$' -count=1 -v -timeout 30m
-
   ssh-first:
     desc: >-
       When a workspace can first be logged into from the host over SSH, through the machine's
@@ -218,28 +204,6 @@ tasks:
     cmds:
       - SPIN_SSH_FIRST=1 go test ./boot/ -run '^TestFirstSSH$' -count=1 -v -timeout 30m
 
-  exits:
-    desc: >-
-      What each phase of a boot made KVM do, from the host: VM exits by kind, EPT faults, and
-      how the guest's RAM is mapped (4K/2M/1G, and how much of it the host holds in huge
-      pages), for the machine as built and without its virtio-rng device, interleaved. Reads
-      /sys/kernel/debug/kvm, so the host must run one VM at a time. Needs KVM, a built release
-      and sudo. REPS= boots of each row, CPUS= vCPUs.
-    deps: [':tools']
-    cmds:
-      - SPIN_BOOT_EXITS=1 go test ./boot/ -run '^TestExitTimeline$' -count=1 -v -timeout 30m
-
-  mmio:
-    desc: >-
-      Who the guest's exits to QEMU are, counted from QEMU's own trace from a launch to the
-      login prompt: every PCI configuration access by device and direction, and every MMIO
-      and port access by the region it reached. Counts, not times, so any host answers.
-      SPIN_KERNEL_B= counts a second kernel beside the release's. Needs KVM and a built
-      release; no sudo. TOP= regions.
-    deps: [':tools']
-    cmds:
-      - SPIN_BOOT_MMIO=1 go test ./boot/ -run '^TestMMIOCensus$' -count=1 -v -timeout 10m
-
   firmware-stages:
     desc: >-
       Where the time before the kernel goes on the machine a workspace boots: QEMU's start and

boot/bench_test.go

@@ -88,8 +88,8 @@ func without(units ...string) variant {
 //     boots, 2026-09-26); image/build.sh forbids the disk.
 //   - serial-getty Type=simple instead of Type=idle: within noise of the echo (2026-09-10).
 //     And again once spin lingered and user@1000.service was in the boot transaction: 346/358
-//     against 423/441, 20 boots, lab run 43362067503 (2026-11-13) - now the image's
-//     (serial-getty@.service.d/type.conf).
+//     against 423/441, 20 boots, lab run 43362067503 (2026-11-13), and in the image for a day:
+//     removed because nothing a workspace does waits for the console's prompt.
 //   - loop.max_loop=0 8250.nr_uarts=1, and CONFIG_VT=n: no change to a login prompt
 //     (2026-09-26). udev's coldplug overlaps the boot rather than delaying it.
 //   - 19 udev rule files for hardware the machine cannot have, masked: no change, same day.