Bug Report: CloudRunner Operation Counting Regression (MAP Guardian v2.3)

Summary:
My MAP Guardian agent, which has been running successfully for months without modification, is now terminating mid-execution without errors. The issue appears to stem from a recent change in CloudRunner’s internal “operation counting” logic.


Details:

  • Agent name: MAP Guardian v2.3

  • Workspace: Personal Workspace

  • CloudRunner setting: Maximum Operations per Invocation = 1506 (increased to 2000 for testing)

  • Expected behavior: Full CSV processing of ~30 product rows (previously completed in ~1200 operations).

  • Actual behavior: Run halts silently after ~13–17 products processed. No “undefined,” “TypeError,” or timeout messages appear.

  • Observed log pattern:

    • Workflow runs normally up to the same index on each attempt.

    • Final log entry shows a standard step (not an error).

    • Next node never executes, indicating CloudRunner termination.

  • Key evidence:

    • CSV successfully parsed (see Query Data Source output).

    • No JSON errors or unhandled exceptions in subsequent blocks.

    • Abrupt stop correlates precisely with estimated 1500–1600 operation usage.


Reproduction Steps:

  1. Open MAP Guardian v2.3.

  2. Load full CSV input (productList ~30 items).

  3. Execute workflow in CloudRunner.

  4. Observe that the run halts mid-iteration with no error output.

  5. Reducing the dataset to ≤10 items allows a successful run.


Analysis:
This appears to be a regression in CloudRunner’s internal operation metering.
Previously, simple variable and JSON access steps counted as 1 operation; now they consume multiple operations each, effectively reducing available runtime capacity by 30–40%.
The behavior began around mid-November 2025.


Request:
Please verify whether CloudRunner’s internal operation accounting changed.
If confirmed, I request either:

  1. Restoration of previous operation behavior, or

  2. Temporary increase of my invocation cap beyond 2000 to restore previous functionality (this agent has no infinite loops or runaway logic).


Supporting Evidence:

  • Run IDs:

    • mindstudio-run-b10f1ae6-6da6-47ee-9a6b-39265b1e622c

    • mindstudio-run-576f354d-1b1c-49d8-934d-3fff41805966

    • mindstudio-run-50ad66d9-1b30-4917-980f-d8250c68f3ea

  • Screenshots and logs attached showing consistent early halts and no explicit errors.


Impact:
This regression prevents MAP Guardian from completing its standard scan, limiting it to ~15 products per run and interrupting automated MAP enforcement workflows that previously executed reliably.

Hi @rdearmas,

Could you share the Debugger logs for those runs? To do that, open the Debugger, select the run, click Share, and reply with the URL: