NOTE: This issue comes from applying #12273 (and the related reva PR). Without those PRs, we aren't tracing events so the problem isn't visible.
Tracing information above doesn't look right. The problem is that the initial search event span finished at 7µs, and there are several related calls (accordingly to the tracing information) that happen after 6 seconds, even longer.
The pattern follows a network pattern where several calls, likely independent, happen during a timeline. The expected pattern should be an inverted pyramid where the initial span holds the whole timeline.
This seems to be caused by how the contexts are being handled by the debouncer. The actual function being executed by the debouncer is delayed, but the context used still refers to the received event.
In addition, the IndexSpace call made by the debouncer isn't reported, so it's difficult to know which is the parent call. There are several sequential calls to the gateway, but no method seems to take credit.
NOTE: This issue comes from applying #12273 (and the related reva PR). Without those PRs, we aren't tracing events so the problem isn't visible.
Tracing information above doesn't look right. The problem is that the initial search event span finished at 7µs, and there are several related calls (accordingly to the tracing information) that happen after 6 seconds, even longer.
The pattern follows a network pattern where several calls, likely independent, happen during a timeline. The expected pattern should be an inverted pyramid where the initial span holds the whole timeline.
This seems to be caused by how the contexts are being handled by the debouncer. The actual function being executed by the debouncer is delayed, but the context used still refers to the received event.
In addition, the
IndexSpacecall made by the debouncer isn't reported, so it's difficult to know which is the parent call. There are several sequential calls to the gateway, but no method seems to take credit.