Split checkpoint guards between VirtualHQ and local defence; record Workshop 0.0.151

docs/configuration.md

@@ -51,9 +51,9 @@ Both counts can be disabled separately with zero; set both to zero to disable al
 
 Roads at or above the width cutoff get large checkpoints; roads below it get compact posts. This is a width classification from the terrain's road network, not a guaranteed asphalt/dirt or forest distinction. An exhausted or disabled quota cannot consume the other type's allowance. Both types share the existing minimum spacing, so they cannot overlap each other.
 
-Small checkpoints retain four guards, four sandbag walls, two approach barriers and two signs (about 17.8 m long). Large checkpoints have eight guards, eight sandbag walls, four approach barriers and two signs (about 42.8 m long), with four protected guard bays on each shoulder. Large candidates require at least 60 m of road and a straight clear approach. Both preserve a straight vehicle passage at least four metres wide.
+Both checkpoint sizes have four guards: two HQ responders and two local sentries who hold the post outside HQ control. Small checkpoints retain four sandbag walls, two approach barriers and two signs (eight props, about 17.8 m long). Large checkpoints retain eight sandbag walls, four approach barriers and two signs (14 props, about 42.8 m long), with four protected bays on each shoulder. Large candidates require at least 60 m of road and a straight clear approach. Both preserve a straight vehicle passage at least four metres wide.
 
-Soldiers come from the selected enemy faction, use normal equipment and physical radio reporting, and register with VirtualHQ at their actual four- or eight-soldier strength. They do not reveal player locations simply by spawning. Guards are not replaced after death. Every post is constructed one actor, prop or waypoint per callback; a failed or interrupted build rolls back without consuming either quota. At maximum settings the two types can create up to 360 guards and 660 replicated props; their default combined count is eight posts and 48 guards. The Status page reports placed/requested counts separately.
+Soldiers come from the selected enemy faction and use normal equipment. Each post registers only its two responders with VirtualHQ; the two local sentries remain outside HQ control. Responders use physical radio reporting and do not reveal player locations simply by spawning. Guards are not replaced after death. Every post is constructed one actor, prop or waypoint per callback; a failed or interrupted build rolls back without consuming either quota. At maximum settings the two types can create up to 240 guards (120 HQ responders and 120 local sentries) and 660 replicated props; their default combined count is eight posts and 32 guards (16 HQ responders and 16 local sentries). The Status page reports placed/requested counts separately.
 
 Placement uses the loaded terrain's road network and Soldiers navmesh after
 terrain/site preparation. The initial search yields after 12 seconds; unfinished
@@ -67,7 +67,7 @@ requested totals instead of preventing mission startup. The admin Status page
 also shows these totals. This is an interim checkpoint system, independent of the
 existing HQ-directed road interceptions.
 
-The [1 November native checkpoint test](evidence/large-checkpoints-2026-11-01/native-checks.log) placed one small and one large post on Everon: 12 living guards, 22 replicated props, correct HQ strength, model footprints outside the centre passage, and separate completed quotas. The [menu/settings fixture](evidence/large-checkpoints-2026-11-01/menu/native-checks.log) saved and reloaded distinct counts and width, and measured the new controls’ native text bounds. It instantiates the generated layout directly; it does not claim human input, remote RPC delivery or screenshot acceptance. Both completed with unchanged inputs and cleanup; strict runners retain 13 existing resource/world errors and no script errors. Earlier failed iterations are recorded alongside the evidence. Full-load performance and multiplayer acceptance remain unmeasured.
+The historical [1 November native checkpoint test](evidence/large-checkpoints-2026-11-01/native-checks.log) used the former four-guard small and eight-guard large layout, before the HQ/local split. It placed one of each on Everon: 12 living guards, 22 replicated props, correct HQ strength for that version, model footprints outside the centre passage, and separate completed quotas. It does not validate the current four-guard split. The historical [menu/settings fixture](evidence/large-checkpoints-2026-11-01/menu/native-checks.log) saved and reloaded distinct counts and width, and measured the controls’ native text bounds. It instantiates the generated layout directly; it does not claim human input, remote RPC delivery or screenshot acceptance. Both completed with unchanged inputs and cleanup; strict runners retain 13 existing resource/world errors and no script errors. Earlier failed iterations are recorded alongside the evidence. Full-load performance and multiplayer acceptance remain unmeasured.
 
 Fresh settings now initialize both drone keys, including the valid 300-second cooldown; the missing cooldown previously prevented a fresh settings document from passing save/readback validation.