Known Issues

FOUND 2026-06-16c while fixing #72: ~500 Mech rows have tonnage != their variant's canonical mass (e.g. Phoenix Hawk/Vindicator stored 50t should be 45t; Corsair 85->95) — #32 contract-reward-spawner straggler family. Surfaces as wrong walk/run on those mechs. FIX READY (api/tmp/fix_tonnage_corruption.rb, APPLY=true): sets tonnage to the canonical loadout mass + weight_class. ~500 mechs; safe (canonical data, #32 precedent). Needs CO greenlight. | 2026-06-16c: APPLIED on prod (CO greenlit) — 350 mechs corrected to canonical mass + weight_class. Re-scan: 0 remaining.

| 2026-06-16c: Server data is CORRECT everywhere — Clan SRM 2/4/6 = 0.5/1.0/1.5t in equipment_catalog, WeaponStatsCatalog, WeaponSlotCatalog AND the /weapons FetchWeapons payload (verified on prod); client DTO is float, palette formats :0.#. Fixed by today's IS/Clan weapon separation. Player likely saw a STALE session — WeaponRegistry fetches the catalog ONCE per session, so anyone logged in before the deploy keeps old weights until relog. RECOMMEND: re-fetch weapons on mech-lab open (small client change, next build). NOT marking resolved until confirmed in a fresh session. | 2026-06-16c CORRECTION + FIX: not a stale session — real CLIENT bug, exactly as CO suspected (and it surfaced AFTER the Clan split). StoreItemDto.tonnage was `int`, so the store truncated the server's 0.5 (Clan SRM 2) to 0 at JSON parse. The split added 'Clan SRM 2' (0.5t) as a distinct store row a Clan faction (CNC) browses, exposing it. FIX: StoreItemDto.tonnage int->float + MarketPanel {:0.#}t (ships next build); also server store.rb legacy-IS path .to_i->.to_f.round(1) (IS MG/small-laser/flamer 0.5t were truncating too — deployed). Verified store now serves Clan SRM 2 = 0.5 / IS MG = 0.5.

| 2026-06-16c: FIXED + deployed. Card/list walk MP now derived from the engine (rating/tonnage), matching the customization menu — was a stale stored base value. Verified on the reporter's Marauder #103655 (75t, engine 380): now derives 5 (was 4). Editor also stores the derived value on commit. (Separate: ~500 wrong-tonnage mechs filed.)

OBSERVED (CO): CNC has captured several DC worlds and DC mounts no counter-offensive against CNC territory. Need to verify how large units / factions react to losing planets and ensure they actually retaliate by attacking the aggressor's planets after losing territory to them. INVESTIGATE: does AiFactionTurn track recent territory losses + the faction that took them? Is there a retaliation/grudge trigger analogous to the pirate-hunt mechanic (raided >=2x -> hunt)? Likely add: on planet loss, record aggressor; AI prioritizes a counter-attack on the aggressor's nearest planet when it has an idle combat fleet in range. Scope to capable factions (fleets + decks available). NEXT SESSION work item. | 2026-06-16c: FIXED + deployed. AiFactionTurn#maybe_retaliate_for_lost_planet (reactive phase, beside the pirate hunt): a faction that lost a world to an aggressor in the last 8 turns mounts a planetary_assault to RETAKE it (declare on-site or dispatch via PendingThreatPursuer). Skips allies; ComStar bulletin per aggressor. Sim-time lookback (valid_until is sim-time). Verified on dev (CGB retook Casere vs CSJ).

ROOT CAUSE (diagnosed via WeaponAttach DIAG): bundle chrprfmech_kitfoxbase-001 ships prefab chrPrfMech_cicadaBase-001 re-skinned with the kitfox mesh + ONLY cicada weapon props. PrefabChassisHint -> 'cicada', so cicada props attach; arm props align but the cicada center-torso missile rack floats in front of the kitfox nose (wrong hardpoint for this mesh). No kitfox props exist in the bundle. Stormcrow = same pattern via cataphract. Likely affects several RogueTech re-skins. FIX OPTIONS: (a) editor hand-position the offending torso prop per affected bundle; (b) supply real kitfox/stormcrow prop sets; (c) interim: bundle-gated suppression of torso props that land far off the torso bone. Broad heuristics regress working rigs (reverted). NOTE: a verbose DIAG line was left in MechWeaponAttacher for auditing. ITEM ID (CO 2026-06-16): the floating prop is chrPrfWeap_cicada_righttorso_srm4_mh1 (Cicada right-torso SRM-4 rack, used for the Streak SRM 4); secondary torso prop chrPrfWeap_cicada_centertorso_flamer_eh1 (LB-5X). CONFIRMED via WeaponAttach DIAG: root='chrPrfMech_cicadaBase-001(Clone)', mechPrefabsInBundle=[chrPrfMech_cicadaBase-001], propChassis=[cicada] — i.e. a KITFOX mesh re-skinned onto the cicada prefab, wearing cicada props. THE FIX (CO direction): there is NO per-chassis override map today — placement is purely the prop name's section + the prop's own authored chrMdlWeap offset (heuristic 'uncollapse'). MechMountExtractor.cs was only PHASE-1 DUMP; the planned Phase-2 'bake' override map was never wired. Right fix = wire that override map: a (chassis|section|hpId) -> position/rotation table the attacher consults BEFORE the heuristic, populated first for chrPrfWeap_cicada_righttorso_srm4_mh1 on the kitfox mesh (capture correct transform in-editor, or extend MechMountExtractor to dump cicada/kitfox). Broad heuristics regress working rigs (reverted). NEXT SESSION owns this. TARGET PLACEMENT (CO 2026-06-16, per Sarna ref): the floating SRM rack belongs on the KitFox's RIGHT ARM (shoulder-mounted launcher). Next session: reposition chrPrfWeap_cicada_righttorso_srm4_mh1 onto the kitfox right arm. CRITICAL CONSTRAINT: the override MUST be KITFOX-ONLY — NOT a global change. The override map keys on (bundle/chassis | section | hpId) precisely so each broken re-skin is fixed SURGICALLY and in isolation; do not touch the shared heuristic that the working rigs depend on. | 2026-06-16c: FIXED in code — MechWeaponAttacher per-(bundle,section) re-skin torso re-route (commit 1fb7b125). KitFox + Stormcrow verified in-editor. Ships in tonight's build.

| 2026-06-16c: FIXED + deployed. Quickdraw is 60t heavy (chassis catalog was already correct; a stray Mech row was spawned at 50t/medium — live-repaired to 60t heavy). No client build needed.

| 2026-06-16c TRIAGE (CO): split. (1) Nova Prime 12 ERML spread over the mech instead of arms = same RogueTech re-skin float family as #69/#19/#22 — fix via ReskinTorsoReroute. (2) Ebon Jaguar Prime loadout is wrong in our data (we just added EJ via #48); canon: LA LB-X 5 (+2t Cluster/Slug), LT LRM-10 (+2t), RT SRM-2 (+1t, +2t Gauss ammo), RA Gauss Rifle — DATA fix. Reporter will file more loadout issues; treat as ongoing loadout-accuracy. | 2026-06-16c: BOTH reported loadout issues FIXED + deployed. (1) Nova Prime 12 ER ML now all-arms (see #19/#22). (2) Ebon Jaguar Prime corrected to canon (Sarna): Gauss/LB-5X/LRM-10/SRM-2/ER-ML; existing EBJ-Primes re-equipped to the corrected canon (9). Seed-side (mech_loadouts.json), survives reset. Reporter said they'll file more loadout issues as new tickets.

| 2026-06-16c: FIXED + deployed. LB-X autocannons have no minimum range (canon). LB 2-X/5-X (IS + Clan) were inheriting AC/2 (4) / AC/5 (3) min range via fallback — set to 0 in equipment_catalog.json. LB 10-X/20-X were already 0. Verified stats + mech-lab both read 0.

REPORTED (CO 2026-06-16): the post-battle VICTORY/AAR modal has a fixed height, so on longer results the bottom 'RETURN TO HUB' button sits on top of the last line (the 'Field salvage … available in Assets -> Salvage' text), hiding it. FIX: size the modal to its content (or scroll the body / pin the button below the content) so the button never overlaps text. Client = AfterActionReportModal. | 2026-06-16c: FIXED in code. End-of-battle panel detail block is now a masked, scrollable ScrollRect — long casualty/salvage lists scroll instead of overflowing onto the RETURN TO HUB button. Ships in tonight's build.

REPORTED (CO 2026-06-16): on maps with a central dip (low elevation or deep water), the top of the floating-island underside mesh was visible through the dipped hexes. FIXED (staged, next client build): BattleV2Bootstrap now tracks the LOWEST hex floor Y placed on the board (incl. negative elevation + sunken water) and hangs the island underbelly's top (rimDepth) just below it with 2.5u clearance, computed dynamically per map. Flat maps (floor >= -1.8u) are unchanged. | 2026-06-16c: Fix staged (BattleV2Bootstrap hangs the island underbelly below the lowest hex). Ships in tonight's build.

| 2026-06-16c DESIGN (CO): NO manual repair. Damaged factories/shipyards auto-recover over time; SURFACE the countdown only. Operations production view: a damaged factory shows status 'Damaged' + 'X weeks (Y hours real time) until fully functional'. Same for shipyards. BUILD: (server) ensure a recovery ETA exists + auto-restore at it (no recovery tick today); (client) show status + dual-time ETA. Bucket: MEDIUM. | 2026-06-16c: FIXED + deployed. Damaged factories/shipyards now AUTO-RECOVER at recovery_eta (NEW IndustryRecoveryCompleter hourly + lazy-restore on next build) — previously they stayed broken forever. Operations now shows 'Damaged · X weeks (Y hours real) until fully functional' (server-computed; real = sim/time_scale). No manual repair, per design. Cleared 6 stuck factories this sweep.

DONE (2026-06-16): Clan Nova Cat had no home 'Mech factories. Added their signature touman on the capital Mualang — Huntsman, Shadow Cat, Warhawk, Mist Lynx, Adder (light through assault). Seeded into db/seeds.rb so it survives a reset. Note: Ebon Jaguar / Nova Cat aren't in the chassis catalog yet (no model), so they're held for a later add.

ROOT CAUSE (client): the per-poll reconciliation only walks an actor one move at a time. When a single poll captured several enemy moves (fast/offline-AI turns), the actor animated move #1 and SILENTLY DROPPED the newer server positions (MoveTo/MoveAlong bail while IsMoving, and the hard-snap fallback required !IsMoving) -> mech appears to 'walk in place' then jumps many hexes later. Firing on the stale rendered hex resolved on the real occupant (an ally). FIX (staged, next client build): MechActor exposes MoveTarget + StopMovement; reconciliation interrupts + hard-snaps any actor mid-walk toward a now-stale hex, bounding visual desync to one poll. Server was always correct (moves are in the turn log). | 2026-06-16c: Client reconciliation fix staged (smooth move-replay; no mid-walk snap to a stale hex). Ships in tonight's build.

| 2026-06-16: Fixed — full IS/Clan weapon separation deployed 2026-06-16. Clan weapons now use their correct Clan weights (Clan ER PPC 6t, Flamer 0.5t, ER Large 4t, LRMs ~half, MGs 0.25t, etc.) in the store, Mech Lab, and combat. Store now gates by tech base.

| 2026-06-16: Server deployed. Apply a build to multiple same-chassis mechs (or one) with a buy-missing-parts flow: a preview returns the missing-parts list + total store cost, you accept, and it buys the parts then applies to every selected mech. (Also fixed a latent bug where saved templates 500'd.) The multi-select UI + accept/cancel cost pop-up ship with the next client build.

| 2026-06-16: Fixed + deployed. Assigning a non-active pilot (the 4 callsigns each had a BONDSMAN variant — bondsmen can't pilot until they earn warrior status) returned an unhandled 500; it now returns a clean error. The assign picker also gains an 'assignable' flag so bondsmen/wounded/on-quest pilots are grayed out (ships next client build).

| 2026-06-16b: Resolved by deployed heat fixes: shutdown is now evaluated on NET heat after heat-sink dissipation (not gross mid-fire), and double heat sinks dissipate at full rate. TimberWolf TBR-D (15 double sinks = 30 dissipation) walking + firing both ER PPCs = 31 gross -> net 1, no shutdown. Please re-test.

Phoenix Hawk read 50t (45t correct) and some Wolverines 50t (55t correct). Fixed 153 affected 'Mechs live; root cause (contract-reward spawner hardcoded 50t/medium) corrected to use the canonical chassis registry.

Likely seen before today's fix. The mixed missile+direct volley fix is deployed — a NARC should now fire alongside your other weapons in one alpha. Please try again and let us know if it persists. | 2026-06-16: NARC reworked 2026-06-16 — it now fires as part of an alpha volley and acts as a missile beacon/magnet rather than blocking other weapons (on top of the earlier mixed missile+direct volley fix). Please confirm it fires alongside your other weapons now.

Plain SRMs already fire unguided (no sensor lock) — only Streak SRMs lock-on, which is canon. If a standard SRM required a lock, please note the 'Mech/weapon so we can repro. | 2026-06-16: Confirmed working as intended — plain SRM 2/4/6 (IS and Clan) fire UNGUIDED (no sensor lock required); only Streak SRMs lock-on, which is canon. Verified post-deploy. If a STANDARD SRM ever demands a lock, note the 'Mech/weapon and we'll repro.

The Clan-tech overlay shipped 2026-06-14 removes the minimum range on Clan autocannons/LRMs and corrects Clan weapon ranges in both the Mech Lab and combat. Please confirm the specific weapon reads right now. | 2026-06-16: Fixed by the tech separation — Clan LRMs and the Clan UAC/5 now have NO minimum range; the Clan UAC/2 keeps min range 2 and Gauss/HAG/artillery keep theirs (per TechManual). Please confirm the specific weapon now reads right.

Found it: jumping was being charged terrain + elevation movement cost, so a jump onto higher/rough ground could exceed the jump budget. Fixed — jumping now costs 1 jump-MP per hex and flies over terrain/elevation. Ships in tonight's release. | Shipped in the 2026-06-15 client build (live).

Same cause as the hired-pilot fix: newly hired pilots weren't available to assign until relog. Fixed — they now appear in the assign picker right away.

Brief gateway blips line up with server deploys (the API restarts for a few seconds). We're batching deploys to off-hours to avoid interrupting play.

Same cause as the sphere-wide heat fix: double heat sinks were dissipating at half rate. Clan 'Mechs (incl. Nova/Adder) now cool correctly. Live.

Confirmed cause: weapon mount locations are correct in our data + the server maps them right; the battle weapons panel only falls back to a name heuristic (lasers -> torso) when a 'Mech's battle state was built from a loadout without per-weapon locations. Cosmetic only — doesn't affect firing or hit locations. Scheduled with a combat-data pass. | 2026-06-16c: FIXED + deployed. Nova Prime canon loadout corrected to 6 ER ML per ARM (was split 3+3 arms / 2+2+2 torsos) — derives arms 6 energy HP / torsos 0, lasers now all on the arms with no torso float. 0 customized Novas reset to canon.

| 2026-06-16c: FIXED + deployed (same fix as #22). Nova Prime's 12 ER ML are now all arm-mounted, so no stray torso/chest barrels. Verified on dev (paper-doll + 3D), CO-approved.

| 2026-06-16c: Dupe of #69 — same RogueTech re-skin torso float (KitFox). Fixed by the re-skin re-route; ships in tonight's build.

Fixed: hire now refreshes the player's roster (heartbeat self-heal + instant local insert next build). No relog needed.

Fixed — ships in tonight's client update. | Shipped in the 2026-06-15 client build (live).

Fixed — ships in tonight's client update. | Shipped in the 2026-06-15 client build (live).

Fixed — ships in tonight's client update. | Shipped in the 2026-06-15 client build (live).

Fixed — ships in tonight's client update. | Shipped in the 2026-06-15 client build (live).

Fixed — ships in tonight's client update. | Shipped in the 2026-06-15 client build (live).

Warnings are now throttled.

Your committed lance now deploys with them.

Fixed a combat desync.

Open a coalition and budget allies in directly.

Parts and raw-materials purchases go through cleanly.

Defense floor now also applies to the AI's offensive muster path (ad-hoc strike forces) — each world keeps a minimum garrison. Staged for the next release.

Clan Flamers, Ultra ACs, ER lasers and LRMs now use proper Clan-tech stats.

Double heat sinks were dissipating at half their rate — now correct.

Stuck fights now resolve on schedule.

Fixed — tooltips now follow the cursor correctly at any resolution. Ships in tonight's client update. | Shipped in the 2026-06-15 client build (live).

FIXED (staged for next client build): detail scroll content pivot was (0.5,1); changed to (0,1) top-left + explicit sizeDelta(0,0) to match the working weapons/event-log scroll content. BattleV2Bootstrap.cs ~line 4915.

| 2026-06-16: Investigated. Server-side the AI advances + fights normally on '16x17 Moorlands 6' (reproduced: all mechs leave spawn, attackers move 12-16 hexes toward the enemy, battle resolves ~turn 4). The board is terrain-clean — no impassable hexes or buildings; the water is a band along the right edge, not a 'round' feature. So the 'AI frozen in the upper-left corner' is the client move-reconciliation desync (see T-0042) under the lag you noted — the AI moved, the client just didn't render it. The 'round construct' has no gameplay terrain behind it (likely a cosmetic forest-grove/prop render). Both should clear with the next client build; if the AI still looks frozen or the round object persists after that, a screenshot + battle ID will let us dig in.

| 2026-06-16b: Likely a client-side 'ghost image' from the move-reconciliation desync (see T-0042) under the lag you noted — the server resolves victory only when a side has no live mechs, so the second Atlas was probably a stale render. The reconciliation fix is staged for the next client build. If a battle still ends in victory with a genuinely live enemy after that build, please note the battle/ID so we can pull the turn log.

INVESTIGATE: engine selector should reflect what the faction actually has (owned in PartInventory + buyable, tech-gated) rather than the full catalog. Check (a) the mech-lab engine-options source (client dropdown + the catalog/options endpoint feeding it), (b) whether engine inventory is read with the correct tech-specific part keys (e.g. 'XL Engine 210' vs 'Clan XL Engine 210' - same key family as the just-fixed Clan engine-save bug), (c) whether the UI should show owned/affordable engines distinctly. Related: MechLab::Editor engine part labeling + PartInventory engine entries.

REPORTED (user feedback): an UNSUCCESSFUL Discord login crashes the client instead of failing gracefully. Repro paths to check: user cancels/denies the OAuth consent, the auth callback times out, or the Discord account isn't linked to a faction. Expected: surface a clear error via the Growler and return to the login screen (no crash). Investigate the Discord OAuth login flow's error/exception handling on the client (callback handler + the unauthenticated/failure branch).

OBSERVED (CO): opening a battle's After-Action replay, mechs appear in the CENTER of the map on the first frame, then snap to their actual starting hexes. Initial replay frame should place each unit at its true start hex (looks like the first state-frame positions aren't applied before the first render / a default center spawn is used). Investigate the replay initializer / first-frame hex placement in the AAR playback path.

REPORTED (tester, 2026-06-16) — PURELY VISUAL. Gameplay is correct: the engine already resolves a real hit component per shot (HitLocationResolver returns a `section`: center_torso / left_arm / head / left_leg ...). The CLIENT just lands every impact VFX at the target's transform origin (root ~= ground/pelvis), so it reads as 'aimed at the crotch'. FIX (client): map each shot's resolved section to the matching body bone/position on the target model and play the impact there — accurate, per-component visuals. Client = BattleV2Bootstrap weapon VFX + MechActor; section data already arrives in the action result hits[].

FOLLOW-UP (CO 2026-06-16): headshots appear to happen more often in allied (coalition) battles — may be coincidence. To investigate: (1) check whether compounded unit bonuses (cohesion / keshik clustering / faction persona / pilot affinity) stack in allied battles in a way that raises to-hit or called-shot/head-hit odds vs solo battles; (2) ADD A HEADSHOT-RATE TRACKER (head hits / total hits, split solo vs allied) so we have data before deciding to tune. Engine: HitLocationResolver (location table + called-shot), Tactical::Engine to-hit mods.

REPORTED (CO 2026-06-16): in a tactical battle, tapping the new SELECT-ALL weapons toggle to check every ready weapon, then pressing SPACE (the FIRE key) skips/ends the player's turn instead of firing the alpha strike. Expected: SPACE fires the full selected volley. To investigate (client): SELECT-ALL likely toggles the weapon checkboxes without arming the fire action, so the SPACE handler falls through to the end-turn/skip path. Check the battle input handler's SPACE binding vs. the SELECT-ALL state, and whether a target/confirm is required.