# Alchemy 429 lines in mc5.log (#1179 #1181 #1183): one cause, one fix (11 Oct)

All three items are the SAME log: firstmin's mc5.log (mirrored to main and research-box). 97 FEED_SRC_DOWN lines in the log tail: 55 alchemy, 42 drpc.

Cause: fm_mc5_live.py line ~600 builds the feed fallback list SRC = drpc, public, alchemy (fm_sim.FAST). The Alchemy key is dead/capped (Alchemy was removed from the team on 8 Oct): every call returns HTTP 429 "billing or cap". The feed tries it each time drpc/public are slow, is told "skipped 60 s", and tries it again after 60 s, roughly once a minute all hour (the 44/38/45 lines). It never works, so it only adds a failed 4 s-timeout attempt to a slow round. Not a duplicate caller on the research/main boxes: they only read the mirror.
Also seen: drpc round >5 s flips the feed to public every ~5 min (01:09, 01:14, 01:19, 01:24): a separate drpc latency issue, not rate limiting.

Fix at the cause (exec; firstmin is live, I do not touch it): in /opt/sw/fm_mc5_live.py replace
  SRC = [x for x in (("drpc", DRPC), ("public", fm_sim.RPC), ("alchemy", fm_sim.FAST)) if x[1]]
with
  SRC = [x for x in (("drpc", DRPC), ("public", fm_sim.RPC)) if x[1]]
(the comment on that line already says Alchemy is last resort only and refuses getLogs over 10 blocks). Restart only when flat (no open trade). Alternative: a working Alchemy key, but Ben removed it on purpose.