diff --git a/TRACKED_ISSUES.md b/TRACKED_ISSUES.md
index 302ea7e7..c16fd491 100644
--- a/TRACKED_ISSUES.md
+++ b/TRACKED_ISSUES.md
@@ -16,6 +16,7 @@ Last updated: 2026-04-23 (v3.0.5 shipped, Max 20x upgrade, three new CC issues f
 
 | # | Title | State | Our involvement |
 |---|-------|-------|-----------------|
+| [#63147](https://github.com/anthropics/claude-code/issues/63147) | Extended-thinking session permanently wedges (`thinking:""`+signature re-sent on disk reload) | Open | Our **#63172 consolidated into it** (canonical "Cluster 13A"). Posted **binary-confirmed mechanism** 2026-05-29 as cnighswonger ([comment](https://github.com/anthropics/claude-code/issues/63147#issuecomment-4574730571)): env-var gates (`CLAUDE_CODE_DISABLE_THINKING=1`/`MAX_THINKING_TOKENS=0` disable thinking; `DISABLE_INTERLEAVED`/`ADAPTIVE` don't) + in-memory-vs-disk trigger taxonomy. Drove cache-fix #162 (opt-in proxy mitigation) + #160 (session-health). CVP-sanctioned binary analysis. |
 | [#34629](https://github.com/anthropics/claude-code/issues/34629) | Prompt cache regression in --resume (~20x cost) | Closed | Root cause analysis, interceptor fix. Original bug that started this work. |
 | [#40524](https://github.com/anthropics/claude-code/issues/40524) | Conversation history invalidated on subsequent turns | Closed | Image persistence discovery, fingerprint analysis, Renvect collaboration. Multiple posts. |
 | [#42052](https://github.com/anthropics/claude-code/issues/42052) | Max 20x plan: 100% usage after 2 hours | Open | Bidirectional TTL data, overage mechanism analysis. TigerKay1926 has contradicting data (stuck 5m TTL). Vergil824 confirmed npm vs standalone cache difference, shared 1h cache patch — pointed to our interceptor (2026-04-09). |