# GMGN 429 follow-up (#2455), testmgr 14 Oct
- Fixes from #2382 are in: CAP 8, holders_snap reserve 6, caller trace /opt/common/gmgn_calls.log (main).
- Trace (8 h): only ~110 limiter-approved calls (rhfast, runtrack, gmgnstream). Yet 111 BAN lines: runtrack 82, gmgnstream 17, rhfast 12. Each approved call right after a ban expires is refused again (consec 2-5, tokens 0).
- So the 429s are NOT from call rate on main (a few calls per 5 min). The server-side ban outlasts our "until" because something outside the limiter keeps hitting the same key. Prime suspect: firstmin's gmgnd (quotes and swaps, see /opt/fm_mirror/gmgnd.log) or any other host using the key; main's limiter cannot see them.
- runtrack is the victim: most frequent probe, so it logs most bans.
- Fix at cause (exec/ops; I cannot touch firstmin): confirm whether firstmin shares this GMGN key/IP; if yes, give runtrack/rhfast/gmgnstream their own key or move them off GMGN. Meanwhile, after a BAN wait the full until + 120 s before the next probe, since each early probe renews the ban.