# #3184 firstmin 429 (public RPC) 18 Oct
Evidence (/opt/fm_mirror, log times BKK): mc5 407 lines total, wl_wlink1 69, wl_cleane 5. Last-hour errors cluster at :48 and :03 (14:48:14/15, 15:03:06/12 hit mc5, wlink1, cleane within seconds) = same 15-min rhythm across all bots. "feed: both RPCs failed" = primary AND fallback both 429 at the same moment.
Cause: bots poll the same public nodes in lockstep, sharing one egress IP, with no jitter and no shared throttle across processes on firstmin. No errors after 15:04 in the mirror; the loop recovers next cycle, no missed trade seen.
Fix (firstmin side, exec only): per-bot start jitter (0-20s) on the 15-min cycle and an rpcpool-style shared slot (SHARED_GAP 0.12) across bots. I cannot touch firstmin. Research-box side is not the source (different IP; twgap already paced in #3179).