It'd definitely be desirable, but it wouldn't be easy. Currently the thing that makes the tool uncomplicated is that it doesn't need to talk to Zotero at all during the scan. All the citation data is added when setting a citation style in LibreOffice. The scan just converts the markers to LO Reference Marks with Zotero format and Zotero item.uris that allow for updating.

The relevant function is here: https://github.com/Juris-M/zotero-odf-scan-plugin/blob/master/chrome/content/rtfScan.js#L271

Hope you like regular expressions ;)

The talking to Zotero bit isn't really too hard. Is https://github.com/Juris-M/zotero-odf-scan-plugin/blob/master/chrome/content/rtfScan.js#L512 the central function that orchestrates the finding and replacing, and https://github.com/Juris-M/zotero-odf-scan-plugin/blob/master/chrome/content/rtfScan.js#L594 the part that does the actual replacements?

If I may ask, why use regexes when FF has XML/XPath functionality built in?

I don't think there's a strong reason to use regex over XML except that Frank likes regex (the original tool this is based on was in python I think, but it's not like that would have made using XML/XPATH impossible). Might be that it actually ends up being more stable given different interpretation of the ODF XML model, but also possible that the reverse is true. Certainly worth testing out.
That looks right wrt the functions, yes.

If I may ask, why use regexes when FF has XML/XPath functionality built in?

You're not the first to ask that question. 😃 The code was originally rejected for inclusion in Zotero for exactly that reason.

The problem is that the target string may be cross-nested with XML tags that capture a larger run of document text. Identifying the string and isolating it for replacement using XML methods would be very hard to do. It would also be slower to run (because you would need to iterate to the top of the XML hierarchy to determine that a given match attempt had failed). I offered that explanation at the time, and it didn't find favor, but that's the reason behind using regex there.