---
name: hfenduleam-target-bloat-ci-red
description: hfenduleam dev target bloat can fill disk → LNK1318 CI RED that mimics a code regression
metadata: 
  node_type: memory
  type: project
  originSessionId: e351ca96-895b-4af5-b548-a9905d84fcd0
---

On hfenduleam (Windows CI runner + doyle's box), the cargo `target/` dir can bloat huge (seen: 396GB, leaving ~2GB free). When the disk fills, the Windows linker fails with **LNK1318 / os-error-112** — which surfaces as a **CI RED that looks like a code regression but isn't**. Fix: `cargo clean` on the runner, re-run → passes.

**Why:** happened 2026-06-13 — M9 Wave 2 PR #8 went RED on the Windows runner; root cause was disk-full, not code (reconcile fix 77e0fb1 was unaffected; re-ran green after clean).

**How to apply:** before treating a Windows-runner RED (especially LNK1318/os-error-112) as a code defect, check free disk / target size; a periodic `cargo clean` on the self-hosted runners prevents it. Don't restart/revert code on this signal alone. Related: [[ci-selfhost-migration]] (runners live on kitsubito + hfenduleam).