docs: write down which boxes are disposable (target selection by blast radius)

Nothing in the repo said which machines are safe to break. The host table gave
access and role and stopped there, so a session needing a victim had to guess --
and the guessing inverted: the two boxes that exist to be broken were treated as
sacred, and DooPlex (the recovery chain) got used because it was the only box no
spec had fenced.

New documentation/runbooks/target-selection.md -- one page, three tiers, and per
machine what is freely permitted / needs care / forbidden, each carrying its
REASON so a rule can be correctly narrowed later instead of ossifying. States the
selection rule positively (start at Tier 0; a Tier 2 box only when a task says so
explicitly; an absent fence is not permission) and that fences name ACTS, not
machines -- demo-hp's over-subscribed local-lvm is one dangerous storage, not a
dangerous box.

CLAUDE.md: host table gains a Blast radius column, gains the missing demo-hp row
(it was where the drill VMs ran and it was not in the table at all), and a pointer
line to the new runbook.

CORRECTION to the spec's problem statement: the designation was not missing. The
2026-07-25 operator ruling naming the t740 as drill+build VM host -- explicitly
"moved off DooPlex" -- already existed in operations/nodes.md. It sat where no
session reads at start, while the prohibitions were repeated in every task spec.
The defect is reachability of the ruling, not its absence, and the R-116 drill on
DooPlex contradicted a written ruling rather than filling a vacuum.

CORRECTION to the R-116 record, same commit: the baseline claimed controller
0.186.0 on both demo boxes. Only felhom-pve was sampled and generalised; demo-hp
re-checked directly runs 0.185.1, so the fleet is split and R-114's TargetAbsent
branch is absent from demo-hp. Fixed in the audit table and REPORT-r116-diag.

Docs only -- no code, no build, no deploy, no host reconfigured, no host renamed.

documentation/audits/DIAG-r116-disks-payload-2026-07-30.md

@@ -14,15 +14,21 @@ All confirmed this run, not carried from the brief.
 | Artifact | Version | How confirmed |
 |---|---|---|
 | agent | **0.115.0** | `felhom-agent --version` on felhom-pve; `felhom-agent/CHANGELOG.md:1` = `## v0.115.0`; repo HEAD `a58239f6ded367fe27b4eb57cf2fcb0c54414b80` |
-| controller | **0.186.0** ⚠ | `pct exec 9201 -- docker ps` on felhom-pve → `felhom-controller:0.186.0`; `CHANGELOG.md:3` = `### v0.186.0`; HEAD `b331f18424b7cf8d0883e7709f74595808264080` |
+| controller | **0.186.0 on demo-felhom, 0.185.1 on demo-hp** ⚠ | `pct exec 9201 -- docker ps`: felhom-pve → `felhom-controller:0.186.0`, demo-hp → `felhom-controller:0.185.1` (both re-verified 2026-07-30); `CHANGELOG.md:3` = `### v0.186.0`; HEAD `b331f18424b7cf8d0883e7709f74595808264080` |
 | hub | **0.81.0** | `manifests/hub.yaml:128` pin **and** the live pod image (`kubectl -n felhom-system get pod hub-… -o jsonpath=…`) both `felhom-hub:0.81.0`; `hub/CHANGELOG.md:1` = `## v0.81.0` |
 | host-install | **1.22.0** | `scripts/felhom-host-install.sh:187` `SCRIPT_VERSION="1.22.0"` |
 | `felhom.eu` HEAD | `c3ce4c7b20ce4c0d3673590d710d13e5d28b68bb` | `git rev-parse HEAD` |
 
-⚠ **The brief's `controller 0.185.1` is not what is deployed.** 0.185.1 is what the **golden bakes**
-(so it is what a fresh drill box installs, which is why Part 5 correctly recorded it); **0.186.0** is
-what both demo boxes actually run, shipped 2026-07-29 as R-114 + R-112. Both numbers are true of
-different things; the distinction matters because R-114's `TargetAbsent` branch exists only in 0.186.0.
+⚠ **The controller version is not one number.** 0.185.1 is what the **golden bakes** (so it is what a
+fresh drill box installs, which is why the brief and Part 5 both recorded it); **0.186.0** shipped
+2026-07-29 as R-114 + R-112 and is what **demo-felhom** runs — but **demo-hp is still on 0.185.1**, so
+the fleet is split. The distinction matters because R-114's `TargetAbsent` branch exists only in 0.186.0,
+i.e. **demo-hp does not have it**.
+
+> **Correction, 2026-07-30.** As first written this row said 0.186.0 was what *both* demo boxes run.
+> That was wrong: only felhom-pve's guest was sampled, and the claim was generalised to the fleet from
+> one box. demo-hp re-checked directly → `felhom-controller:0.185.1`. Corrected in the same commit that
+> added `runbooks/target-selection.md`.
 
 Hub host table, read this run: `demo-felhom-8363b5` agent 0.115.0 ONLINE · `demo-hp-bb76ea` agent
 0.113.0 ONLINE · `drill-r50-0a4f9a` DOWN · `sess-c-8158e4` DOWN · `sess-d-0452c4` **STALE**.