Figured it out!
The issue stems from how Orion uses conventional ID numbers for most objects, but for Custom Node Pollers it uses what looks like unique identifiers (such as 01d8de05-4b07-4d07-98bc-32349f7820b4). Thus, when it tries to use one in a comparison it bombs out. To resolve this, I wrapped each Join On in a ToString when it comes from another table:
SELECT tolocal(AStat.TriggerTimeStamp) AS [Time of Alert], AStat.ObjectName AS [Alert], ADefs.Name AS [Category], tolocal(AStat.AcknowledgedTime) AS [Acked Time], AStat.AcknowledgedBy AS [Acked By], AStat.Notes, CASE WHEN AStat.ObjectType = 'Node' THEN ('/Orion/NetPerfMon/NodeDetails.aspx?NetObject=N%3a' + ToString(Nodes.NodeID)) WHEN AStat.ObjectType = 'Volume' THEN ('/Orion/NetPerfMon/VolumeDetails.aspx?NetObject=V%3a' + ToString(Volumes.VolumeID)) WHEN AStat.ObjectType = 'Hardware Sensor' THEN ('/Orion/NetPerfMon/NodeDetails.aspx?NetObject=N%3a' + ToString(Hardware.NodeID)) WHEN AStat.ObjectType = 'APM: Component' THEN ('/Orion/View.aspx?NetObject=AM%3a' + ToString(APMC.ComponentID)) WHEN AStat.ObjectType = 'Custom Node Poller' THEN ('/Orion/View.aspx?NetObject=N%3a' + ToString(CusPol.NodeID)) END AS [_LinkFor_Alert]
FROM Orion.AlertStatus AS AStat
INNER JOIN Orion.AlertDefinitions AS ADefs ON AStat.AlertDefID = ADefs.AlertDefID
LEFT OUTER JOIN Orion.Nodes AS Nodes ON ToString(Nodes.NodeID) = AStat.ActiveObject AND AStat.ObjectType = 'Node'
LEFT OUTER JOIN Orion.Volumes AS Volumes ON ToString(Volumes.VolumeID) = AStat.ActiveObject AND AStat.ObjectType = 'Volume'
LEFT OUTER JOIN Orion.HardwareHealth.HardwareItem AS Hardware ON ToString(Hardware.ID) = AStat.ActiveObject AND AStat.ObjectType = 'Hardware Sensor'
LEFT OUTER JOIN Orion.APM.Component AS APMC ON ToString(APMC.ComponentID) = AStat.ActiveObject AND AStat.ObjectType = 'APM: Component'
LEFT OUTER JOIN Orion.NPM.CustomPollerAssignment AS CusPol ON ToString(CusPol.CustomPollerAssignmentID) = AStat.ActiveObject AND AStat.ObjectType = 'Custom Node Poller'
WHERE Acknowledged = 1
ORDER BY TriggerTimeStamp
As far as I can tell, this would only affect you if you currently have an alert that is based off of a Custom Node Poller, or on any object that uses a identifier like 01d8de05-4b07-4d07-98bc-32349f7820b4.
Thank you for your help; looking into that log got the gears turning in my head and got me thinking along the lines of "Hey, maybe the data isn't sane ... somehow?".