Uploading scanned document results in it's preview being blank. This happens for all scanned PDFs, i.e. files that have only one image embedded. Other documents are loaded just fine.

Other thing I noticed is that after running the document trough OCR (tesseract) and uploading the generated PDF, the preview looks the same but the embedded invisible text layer is highlightable. So it seems like nuxeo is not fetching right image data from the PDF into pdf.js preview.

This also happens on clean Nuxeo instance (tested on 8.10, 9.2 and 9.3).

Any idea what might cause this and how to fix it?