Something funny happened with the change in
When converting sourceMimeType image/tiff we need to match 'image/tiff' and 'image/*'. There are two converters in nuxeo-thumbnail for image/tiff (pdfAndImageToThumbnail and anyToThumbnail) and there is a converter for 'image/*' in nuxeo-platform-convert called image2pdf
In MimeTypeTranslationHelper the 'image/tiff' converters are fetched and placed in cos.
This causes the converter with image/* to be skipped because cos is not empty. The end result is that no image can be exported as a PDF from the Web UI.
The algorithm should be reworked to return converters matching the specific mime type and the wildcard.