Windows: removing CLAUDE_CONFIG_DIR while claude.exe runs pins one CPU core indefinitely

Finding: In an external report on Claude Code 2.1.292 (native claude.exe, Windows 11 ARM64, interactive ConPTY session), removing the directory named by CLAUDE_CONFIG_DIR while the process runs reportedly makes it spin at about 1.1 CPU cores indefinitely, with no error shown and flat memory (https://github.com/anthropics/claude-code/issues/101027). This is an external report; KFA did not reproduce it. Reported measurements, each a 30 s CPU delta: config dir present and populated, 0.002 cores; contents deleted but directory kept, 0.001 cores; directory itself removed, 1.109 cores. The reporter includes a Node 22 + node-pty script for the sequence. In the field, leftover processes reportedly ran 15-42 h at 0.7-1.0 core before being killed by hand (same issue). Cause: unknown. The reporter hypothesizes the bundled chokidar add() re-adds a path or its parent with no backoff once the directory is gone, but states this is unverified and that no stack or I/O counters were captured. They note similarity to issues #76608, #81222 and #100831, none of which is an exact duplicate (same issue). Second, related report: claude.exe may stay alive after its console/ConPTY closes, especially when a dialog, tool call or background task is active and ignores a typed exit sequence. The reporter says this came from field observation and was not isolated in a controlled repro. Maintainer status: the issue is open and no maintainer response or fix is present in the supplied content. No fix commit or release is linked. Reporter-proposed mitigation for terminal hosts (not verified by KFA): do not delete a session's CLAUDE_CONFIG_DIR until its claude.exe is confirmed dead; emptying the contents did not trigger the spin in their test. On pane close, kill the actual claude.exe process after a short grace period, since killing a wrapper root process may not be enough. Send Esc before the exit sequence so open dialogs do not swallow it. Limitations: single report, only the issue head was available (no comments), and the effectiveness of the mitigation was not independently confirmed.