A stalled reply says it is asking again while Escape only goes Home

Start with internal/session/loop.go:2694 and :2471, then inspect internal/tui3/failurerow.go:56 to trace both retry notices and Escape handling. Reproduce through codeaf chat in tmux with the provided stub.py, then add the requested unit and e2e coverage so the displayed promise and request state match Escape's actual effect; update the retry manual and invalidates.

Found on santos/dev2 at 008363c98 (#1410). It reaches dev when #1410 merges.

What happened

On a mid-reply stall, the person saw the model went quiet mid-reply — asking again. The documented wait line no answer N in T · still asking · esc stops was never drawn for that shape. In a conversation, Escape went Home while the re-ask continued in the background. The re-ask behavior is new relative to dev, whose corresponding hand test gave up after two attempts.

Replication

Deterministic (no model). A stub provider that sends one content chunk and then holds the connection open. Save it as stub.py:

# stub.py: a provider whose every reply is one content chunk.
#   python3 stub.py cut   -> then the connection closes (no finish_reason, no [DONE])
#   python3 stub.py hold  -> then the connection stays open for ten minutes
import json, sys, time
from http.server import ThreadingHTTPServer, BaseHTTPRequestHandler
MODE = sys.argv[1]
MODEL = "deepseek/deepseek-v4-flash"
class H(BaseHTTPRequestHandler):
    protocol_version = "HTTP/1.1"
    def log_message(self, *a): pass
    def do_GET(self):  # the model listing
        b = json.dumps({"object": "list", "data": [{"id": MODEL, "object": "model"}]}).encode()
        self.send_response(200); self.send_header("Content-Type", "application/json")
        self.send_header("Content-Length", str(len(b))); self.end_headers(); self.wfile.write(b)
    def do_POST(self):
        self.rfile.read(int(self.headers.get("Content-Length", "0")))
        self.send_response(200); self.send_header("Content-Type", "text/event-stream")
        self.send_header("Connection", "close"); self.end_headers()
        chunk = {"id": "c", "object": "chat.completion.chunk", "model": MODEL,
                 "choices": [{"index": 0, "delta": {"content": "Partial ans"}, "finish_reason": None}]}
        self.wfile.write(("data: " + json.dumps(chunk) + "\n\n").encode()); self.wfile.flush()
        if MODE == "hold": time.sleep(600)
        self.close_connection = True
srv = ThreadingHTTPServer(("", 0), H)
print("http://0.0.0.0:%d/v1" % srv.server_address[1], flush=True)
srv.serve_forever()

export CODEAF_HOME=$(mktemp -d) HOME=$(mktemp -d) OPENROUTER_API_KEY=<any non-empty value>
U=$(mktemp); python3 stub.py hold > "$U" & sleep 1
R=$(mktemp -d) && git -C "$R" init -q && cd "$R"
CODEAF_BASE_URL="$(cat "$U")" codeaf chat --no-host

Send reply with the word OK and wait about a minute. Today the row reads the model went quiet mid-reply — asking again and the request is sent again roughly every 47 s; no answer … still asking · esc stops is never drawn. Press esc: Home opens and the stub keeps receiving requests.

Field (real models). A real provider stall does the same, but it cannot be summoned; the stub is the replication.

Where

internal/session/loop.go:2694 emits the quiet-mid-reply notice; :2471 formats the distinct no answer … esc stops line. internal/tui3/failurerow.go:56 draws the retry event.

The fix

Make the visible escape promise match the actual key behavior on every retry shape. If Escape is meant to stop this wait, route it to the active request before navigating Home; otherwise remove esc stops from the wait line and document background continuation.

Acceptance

• e2e: through codeaf chat in tmux against a stalled stub, the displayed line accurately predicts Escape's effect, and the request state follows that effect.
• Unit: tests cover both notice strings and the active-turn Escape route.
• Update the retry manual and invalidates.