This is not how it is done. Generally the writing uses the underlying morphemes, and pronunciation is then fitted on after the fact (but indeed the pronunciations aren't arbitrarily fitted, there are regular sound correspondences that account for > 95% of the words). This is why the 荔鏡記 can be written (and why other variety-specific novels have also been written using the same script over the centuries). Indeed the simple fact is that variety-specific works have existed since at least the Ming Dynasty (and there are quite a few as soon as we hit the Qing Dynasty) and they have been written with the same script.
This is also why it's not a syllabary (and also why the varieties actually have a lot of similarities). This works precisely because varieties share an overwhelming number of cognates if not just straight up the same words.
It can and has, even historically (see e.g. the Ming Dynasty Min novel Tale of the Lychee Mirror 荔鏡記). The reason it falls flat is the lack of political willpower to really go for it, not fundamental linguistics ones. This is why calling Chinese characters a simple syllabary is also wrong.
The relationship among different Sinitic varieties is more complex than simply different languages. I've talked about this in the past: https://news.ycombinator.com/item?id=16844074