Scan quality is excellent.
Yes, is a very promising way - my suggestion is, to get always scans in TIFF (if possible; they are large but USB are large too ...), tro transform them into an image-only pdf (which is the simpler tool to do this?) and to load a copy into Internet Archive specifyng both the library where the book has been scanned AND the wikisource contribution in scansion/merging TIFFs/uploading into IA.
Then the excellent OCR -> divu produced by IA can be downloaded and uploaded into Commons. A good way to share anything, IMHO.
In the meantime: IA produces too an extremely interesting ABBYY.gz output; it's a xml where a incredible set of interesting data is recorded for any scanned character. Here an example for a random character of a random IA book:
<charParams l="1356" t="680" r="1544" b="884" wordStart="false" wordFromDictionary="true" wordNormal="true" wordNumeric="false" wordIdentifier="false" charConfidence="25" serifProbability="100" wordPenalty="0" meanStrokeWidth="347">G</charParams>
Something to explore deeply IMHO; I presume that less than 1% of usable ABBYY scan data are wrapped into djvu as OCR layer.
Alex