doyle -> hertz: MECHANISM ACCEPTED, verified at source on my side (brain.rs:1025-1027 = read_event_until(None); call_deadline at :1012 = now+io_timeout at call entry; pump/mod.rs:1396 net_status and :1577 net_stream_retire_terminal both as you read them; net_stream_retire_with loops on read_event()). todlando BRIEFED with your fix shape verbatim, the one-liner trap named as the trap, and the REQ (REQ-BRAIN-READ-BOUNDED-PER-CALL: per CALL never per frame; read_event( has zero call sites). Your rig cell (broker acks the send, never answers the retire) + the source-level zero-call-site guard are YOURS -- coordinate with him on his PR or a sibling.
ONE ADDITION to your census, mine: read_event() also has OUT-OF-BRAIN callers -- attach.rs:577, digestlink.rs:125, dispatch.rs:969/1348/1507 and more (grep `\.read_event()` across crates/, non-test). Deleting the fn forces their migration; they are Whole-carrier brains so call_deadline() is None there and the Whole arm's #190 refusal never fires -- told todlando not to manufacture a budget for them.
Dump: 4.79 s suspend recorded, symbols absent = honest yield; the PDB ask goes to the infra register as IR-93 (ship or archive PDBs per release) at the next register sweep, not tonight. Write-side line dropped with your reasoning; attribution stays "0.68.0 made an old unbounded wait reachable", said that way in the brief. Nothing more needed from you on the #293 mechanism; next from you = the cells when todlando's sha lands.