diff --git a/notes/observed-bugs.md b/notes/observed-bugs.md
index 7c547f423..7fbe8e7df 100644
--- a/notes/observed-bugs.md
+++ b/notes/observed-bugs.md
@@ -82,6 +82,26 @@ without board power-cycle").
 **Status**: task c3 (indirectly, via bes_chardev removal which currently
 gates the signal/nosignal mode switch path).
 
+## Backlog — full architect review of bes2600 driver code quality
+
+The Phase 0 perf trace for Bug #5 exposes a "when in doubt, add a lock"
+pattern in the BH path (~20 % CPU in `_raw_spin_unlock_irqrestore` even
+during healthy throughput). Markus has flagged this for a separate
+architect-review pass: have Claude Sonnet (or equivalent reviewer) do a
+top-to-bottom code-quality review of the bes2600 sources we have on
+boltzmann (`~/src/besser/bes2600-dkms-mobian/bes2600/`), looking for:
+
+- needless lock proliferation
+- BH / workqueue dispatch shape
+- error-handling coverage
+- dead code / leftover-from-cw1200 cruft
+- API contract violations relative to mainline mac80211
+
+Output: ranked list of cleanup targets that would make later patch series
+land more cleanly. Not blocking on Bug #5 — independent track.
+
+**Status**: backlog. Schedule when Bug #5's measurement pass finishes.
+
 ## Bug #5 — RX path degrades under attempted-throughput pressure
 
 **Suspect file**: bes2600 RX path (`txrx.c bes2600_rx_cb`, `bh.c bes2600_bh_work`,