Hello Team, I am currently running a self-hosted instance of O2 (Version: v0.14.7) and am experiencing an influx of false-positive alerts that were not triggering previously .Upon initial investigation, I suspect a data type mismatch downstream. The anomalous behavior appears to stem from incoming metric values being ingested as string literals rather than numeric types. To mitigate this, I implemented a type-casting function within the ingestion pipeline to explicitly convert these values into integers during the parsing phase. Despite this validation logic, the invalid alerting threshold triggers persist. Could you please assist in verifying if there is a known regression in the evaluation engine for this version, or advise on further diagnostic steps?

Update your O2, you are over 80 versions behind.

However, we were not facing this issue before; it only started occurring a few days ago. Since we cannot update the version, could you at least let us know the root cause?

data type of field won't change to int once it has converted to string. This is how schema widening works to avoid any data loss. If you want to have int, then create a new field using VRL and create an alert on that.