fix: reduce default hysteresis 10%→2%; show recovery threshold in alerts UI

The 10% default hysteresis created an unreasonably wide recovery band:
a 95% threshold would only clear once the value dropped below 85.5%,
causing alerts to linger long after the metric was well below the
trigger level.

Change default hysteresis to 2% across all threshold parsers (plugin
metrics, partitions, RTT). For a 95% threshold, recovery is now at
93.1% instead of 85.5%.

Add AlertState.hysteresis field (set on every check, cleared on OK) and
expose recovery_threshold in to_dict() so the Alerts dashboard can
display "recovers < 93.1" alongside the trigger threshold, making the
hysteresis band visible to the user. Pickle backward-compatible via
__setstate__.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

hbd/server/templates/alerts.html

@@ -405,6 +405,10 @@
         } else if (alert.threshold_value !== undefined && alert.threshold_value !== null && alert.operator) {
           valueText += ` <span class="threshold-info">(threshold: ${alert.operator} ${formatValue(alert.threshold_value)})</span>`;
         }
+        if (alert.recovery_threshold !== undefined && alert.recovery_threshold !== null) {
+          const recOp = (alert.operator === '>' || alert.operator === '>=') ? '<' : '>';
+          valueText += ` <span class="threshold-info" style="color:#888">(recovers ${recOp} ${formatValue(alert.recovery_threshold)})</span>`;
+        }
         
         // Build actions section
         let actionsHtml = '';

hbd/server/threshold.py

@@ -57,6 +57,7 @@ class AlertState:
         self.last_notification = None
         self.threshold_value = None  # The threshold value that triggered alert
         self.operator = None  # The comparison operator (>, <, >=, etc.)
+        self.hysteresis: Optional[float] = None  # Hysteresis fraction used for recovery
         self.formatted_message = None  # Formatted display message for UI
         self.acknowledged = False  # Whether alert has been acknowledged
         self.acknowledged_at = None  # Timestamp when acknowledged
@@ -151,7 +152,16 @@ class AlertState:
             result["operator"] = self.operator
         if self.formatted_message is not None:
             result["formatted_message"] = self.formatted_message
-        
+
+        # Compute and expose the recovery threshold so the UI can display it
+        if (self.hysteresis and self.threshold_value is not None
+                and self.operator is not None):
+            ha = abs(self.threshold_value * self.hysteresis)
+            if self.operator in ('>', '>='):
+                result["recovery_threshold"] = round(self.threshold_value - ha, 4)
+            elif self.operator in ('<', '<='):
+                result["recovery_threshold"] = round(self.threshold_value + ha, 4)
+
         return result
     
     def __setstate__(self, state):
@@ -159,6 +169,8 @@ class AlertState:
         self.__dict__.update(state)
         if not hasattr(self, 'consecutive_count'):
             self.consecutive_count = 0
+        if not hasattr(self, 'hysteresis'):
+            self.hysteresis = None
 
     def acknowledge(self):
         """Acknowledge this alert to stop reminder notifications."""
@@ -546,7 +558,7 @@ class ThresholdChecker:
             critical = threshold_config.get("critical")
             operator = threshold_config.get("operator", ">")
             display = threshold_config.get("display", "(threshold: {op_symbol} {threshold_value})")
-            hysteresis = threshold_config.get("hysteresis", 0.1)  # 10% default
+            hysteresis = threshold_config.get("hysteresis", 0.02)  # 2% default
             enabled = threshold_config.get("enabled", True)
             
             if warning is None and critical is None:
@@ -649,7 +661,7 @@ class ThresholdChecker:
         warning = rtt_thresholds.get("warning")
         critical = rtt_thresholds.get("critical")
         operator = rtt_thresholds.get("operator", ">")
-        hysteresis = rtt_thresholds.get("hysteresis", 0.1)  # 10% default
+        hysteresis = rtt_thresholds.get("hysteresis", 0.02)  # 2% default
         enabled = rtt_thresholds.get("enabled", True)
         display = rtt_thresholds.get("display")
         count = rtt_thresholds.get("count", 1)
@@ -794,6 +806,12 @@ class ThresholdChecker:
         elif new_level == AlertLevel.WARNING and threshold.warning is not None:
             threshold_value = threshold.warning
 
+        # Keep hysteresis on the state so the UI can show the recovery threshold
+        if new_level != AlertLevel.OK:
+            alert_state.hysteresis = threshold.hysteresis
+        else:
+            alert_state.hysteresis = None
+
         # Update state and check for changes
         old_level = alert_state.level
         if alert_state.update(new_level, value, threshold_value, threshold.operator.value):
@@ -876,7 +894,9 @@ class ThresholdChecker:
                 threshold_value = threshold.critical
             elif new_level == AlertLevel.WARNING and threshold.warning is not None:
                 threshold_value = threshold.warning
-            
+
+            alert_state.hysteresis = threshold.hysteresis if new_level != AlertLevel.OK else None
+
             # Update state and check for changes
             old_level = alert_state.level
             if alert_state.update(new_level, value, threshold_value, threshold.operator.value):
@@ -942,7 +962,9 @@ class ThresholdChecker:
                         threshold_value = threshold.critical
                     elif new_level == AlertLevel.WARNING and threshold.warning is not None:
                         threshold_value = threshold.warning
-                    
+
+                    alert_state.hysteresis = threshold.hysteresis if new_level != AlertLevel.OK else None
+
                     old_level = alert_state.level
                     if alert_state.update(new_level, value, threshold_value, threshold.operator.value):
                         state_changes.append((metric_path, old_level, new_level, value))