fix(doctor): per-OS compute remedies + resources-set-max drift nudge (#400)

[LukasWodka]

Fixes#400. Companion to tracebloc/client#393 (backend#1236 option A) — the doctor half of the drift story.

What

1. computeRemedy(goos) replaces the unconditional 'raise the machine's allocation in Docker Desktop → Resources' (wrong on Windows's default WSL2 backend — no Resources→Memory slider exists there, field-verified 2026-07-24 — and meaningless on bare Linux):
   • windows: both levers — [wsl2] memory=… in %UserProfile%\.wslconfig + wsl --shutdown, or the Hyper-V slider
   • darwin: keeps the slider (mirrors the existing darwin gate in resources_set.go)
   • linux: no Docker Desktop reference
   • every variant ends with tracebloc resources set max — the shrink-side drift fix. Applied at both emit sites (node-capacity FAIL roll-up + stuck-Pending remedy).
2. Grow-side drift nudge in checkNodeFit's OK path: when the budget uses ≤ half of what the largest Ready node could give one run (node − platform overhead via resources.MaxRunCores/GiB — largest, never summed, resources/home: "equipped with" sums allocatable across k3d nodes — reports 2× the machine (24 CPU · 13.4 GiB on a 12-core host) #399), the detail points at resources set max.

Tests

• computeRemedy per-GOOS invariants; the FAIL-case test now pins the GOOS-independent resources set max instead of "Docker Desktop"; checkNodeFit big-machine⇒nudge / snug⇒quiet (existing OK fixtures verified below the threshold).
• copy-catalog golden regenerated + diff reviewed (only the new remedy strings).

make ci green.

🤖 Generated with Claude Code

---

Note

Low Risk
User-facing CLI copy and an informational verbose detail on an existing health check; no auth, scheduling, or API behavior changes.

Overview
tracebloc doctor now gives OS-correct guidance when compute is tight, and nudges users when their run budget is far below what the machine can offer.

A new computeRemedy(goos) replaces the old one-size Docker Desktop → Resources text (misleading on Windows WSL2 and bare Linux). Windows mentions .wslconfig and Hyper-V; macOS keeps the Docker Desktop slider; Linux avoids Docker Desktop entirely. Every variant ends with tracebloc resources set max. That remedy is wired into the rolled-up “not enough compute” and stuck-pending pod readiness messages.

On the Node capacity check’s success path, when configured CPU/memory requests use at most half of the largest Ready node’s effective ceiling (via resources.MaxRunCores / MaxRunGiB, CPU-major tie-break to match resources set max), the verbose detail appends a grow-side drift hint pointing at resources set max.

Tests cover per-GOOS remedy strings, nudge on/off and heterogeneous nodes; the string golden file was regenerated.

^{Reviewed by Cursor Bugbot for commit b0f235e. Bugbot is set up for automated code reviews on this repo. Configure here.}

[LukasWodka]

bugbot run

[LukasWodka]

bugbot run

[cursor]

✅ Bugbot reviewed your changes and found no new issues!

1 issue from previous review remains unresolved.

Comment @cursor review or bugbot run to trigger another review on this PR

^{Reviewed by Cursor Bugbot for commit 741bc3f. Configure here.}

[LukasWodka]

bugbot run

[cursor]

✅ Bugbot reviewed your changes and found no new issues!

Comment @cursor review or bugbot run to trigger another review on this PR

^{Reviewed by Cursor Bugbot for commit b0f235e. Configure here.}

[LukasWodka]

Functional review \u2014 passed\n\nBasis: the behavioural suites that ran on this PR at merge against real environments, not mocks:\n\n- \u2014 the installer suite exercised for real, plus across all packages\n\nI could not reach the dev API (no dev credentials), so rather than rubber-stamp I used the strongest evidence available: these suites exercise the actual behaviour this change alters, on real infrastructure. Advancing \u2192 .\n\nIf the functional reviewer wants a manual pass on dev in addition, please move it back and say so.