Metric Filter for Lambda Logs Not Reflecting Accurate Memory Usage

I created a metric filter from a Lambda log group to monitor allocated and maximum memory used by the Lambda function. The Filter pattern which is - [...] was previously working correctly, capturing values with placeholders like $13 for allocated memory and $18 for max memory used as suggested here that they are constants. These values accurately reflected memory usage in MB.

An alarm was configured based on this metric with a threshold of >=200 (assuming MB). However, two days ago, the issue began:

Alarm notifications started triggering for memory usage exceeding 6.7554799525E8, significantly higher than the 200 MB threshold. Adding a unit (MB) to the metric resulted in a "unit type is none" error. Now, the displayed metric values are inaccurate: Allocated memory, which is a constant 256MB, spikes to 430 MB in the monitoring graph Max memory used goes even higher. Log stream inspection confirms memory usage never exceeds 200 MB.

Troubleshooting Attempts: Increased the alarm threshold to 200,000,000 (assuming bytes), but it reverted back to 200. I deleted and recreated the metric filters the same way but ran into the same issues Suspect the variable position for the metric value might have changed, but limited output display (only shows $1-$6) hinders confirmation.

Question: Is there a way to view more of the filter pattern output to verify the variable position? If not, what other potential causes could explain the inaccurate metric values?