Sync retry with exponential backoff (15s→15min cap); health exposes retry state

app/main.py

@@ -94,6 +94,10 @@ _state = {
     "last_error": None,
     "sync_attempts": 0,
     "sync_failures": 0,
+    "retry_attempt": 0,          # consecutive failures (backoff exponent)
+    "retry_active": False,       # a backoff sleep is in flight
+    "retry_in": None,            # seconds until the next retry (None when idle)
+    "next_retry_at": 0.0,        # epoch of the next scheduled retry
     "source_used": None,
     "data_updated": None,      # freshest govUK price_last_updated across dataset
 }
@@ -416,6 +420,9 @@ import threading
 
 _refresh_lock = threading.Lock()
 
+_RETRY_BASE_S = 15    # first retry 15s after a failure
+_RETRY_MAX_S = 900    # exponential backoff capped at 15 minutes
+
 
 def _refresh_cache() -> None:
     """Refresh the station cache (called synchronously or from a thread)."""
@@ -447,9 +454,35 @@ def _refresh_cache_locked() -> None:
                 _state["source_used"] = "csv"
         _state["stations_fetched_at"] = time.time()
         _state["last_error"] = None
+        _state["retry_attempt"] = 0
+        _state["retry_in"] = None
+        _state["next_retry_at"] = 0.0
     except Exception as exc:  # keep serving stale cache on failure
         _state["sync_failures"] += 1
         _state["last_error"] = f"{type(exc).__name__}: {exc}"
+        _state["retry_attempt"] += 1
+        delay = min(_RETRY_BASE_S * (2 ** (_state["retry_attempt"] - 1)), _RETRY_MAX_S)
+        _state["retry_in"] = delay
+        _state["next_retry_at"] = time.time() + delay
+        _schedule_retry(delay)
+
+
+def _schedule_retry(delay: float) -> None:
+    """Retry a failed sync on a daemon thread after `delay` seconds.
+
+    Only one retry sleeps at a time; the worker clears the flag before
+    refreshing, so a repeated failure can chain the next (longer) retry.
+    """
+    if _state["retry_active"]:
+        return
+    _state["retry_active"] = True
+
+    def worker() -> None:
+        time.sleep(delay)
+        _state["retry_active"] = False
+        _refresh_cache()
+
+    threading.Thread(target=worker, daemon=True).start()
 
 
 def _ensure_fresh_cache() -> None:
@@ -530,6 +563,10 @@ def health():
         "last_sync": time.strftime("%Y-%m-%dT%H:%M:%SZ", time.gmtime(_state["stations_fetched_at"])) if _state["stations_fetched_at"] else None,
         "sync_attempts": _state["sync_attempts"],
         "sync_failures": _state["sync_failures"],
+        "retry_active": _state["retry_active"],
+        "retry_in": _state["retry_in"],
+        "next_retry_at": (time.strftime("%Y-%m-%dT%H:%M:%SZ", time.gmtime(_state["next_retry_at"]))
+                          if _state["next_retry_at"] else None),
         "last_error": _state["last_error"],
     }