I am calling DocumentAI OCR batch processing from Workflows generally quite successfully, however, I occasionally get the following error:

{
  "caughtError": {
    "message": "An error occurred during execution of a long-running operation.",
    "operation": {
      "done": true,
      "error": {
        "code": 3,
        "message": "Failed to process all documents."
      },
      "metadata": {
        "@type": "type.googleapis.com/google.cloud.documentai.v1.BatchProcessMetadata",
        "createTime": "2024-07-18T12:43:54.456405Z",
        "individualProcessStatuses": [
          {
            "inputGcsSource": "gs://redacted_bucket_name/redacted.pdf",
            "status": {
              "code": 3,
              "message": "Invalid input document content."
            }
          }
        ],
        "state": "FAILED",
        "updateTime": "2024-07-18T12:44:25.393457Z"
      },
      "name": "projects/###/locations/us/operations/###"
    },
    "tags": [
      "OperationError"
    ]
  }

I am able to isolate the error down to the page by breaking the pdf up to see which page is causing the error. However, I am unable to discern what the exact cause of the error is programmatically to fix it.

I am able to use libraries like pypdf and pymupdf to open the pdf and successfully extract the text (if it exists). I have also tried using pymupdf to compress the pdf using various compression settings, and I am able to compress the pdf, but this does not prevent the error. On one pdf where I was encountering the issue, setting enableNativePdfParsing to false allowed the pdf to process, however this does not prevent the error in all cases I am seeing.

Is there some way I can preprocess my pdfs to ensure the pdf is compatible with DocumentAI or is there another way to solve this issue?