Hi Joe,
Yes, total bytes changed would be a relatively easy weighting factor, but not I suspect a useful one. So many edits are semi automated addition of large templates, or reversion of previous changes that you would need to do some complex filtering to identify the "amount of meaningful work". And it isn't just the individual edit that you need to consider, like most active editors I have a number of scripts or tools that I have opted into. In one extreme example, if I use Twinkle to nominate an article for deletion by AFD, all I have to do is click on a couple of menus and type a one sentence case for deleting the article and submit. My account then does the following edits:
1 creates a page for the deletion discussion with various bits of code including one copy of my deletion rationale.
2 Lists that deletion discussion on the page for that days deletion discussions.
3 Templates the article with a warning that it is being considered for deletion, with the rational and several sentences of verbiage.
4 Writes a template to the author's talkpage telling them what I have done, quotes the rational and explains what they can do about it
Note by Wikipedia standards these are four manual edits each adding a generous paragraph or more and all generated by the writing of a rationale that could be shorter than "Not yet played so not yet notable".
For example, I just went to recent changes, by far the biggest edit of that moment was
this one. At first glance it look like someone added 1500 bytes of encyclopaedic text. Then you realise they merely reverted a vandalism of three minutes earlier.
Then there's the issue that some people will go through an article fixing an assortment of typos and perhaps also rephrasing the English, that can be quite a lot of time spent and many changes with little or any total change in bytes. I'm sure it would be possible to come up with some set of filters that compensates for most of this. But it won't be simple and it would always be vulnerable to someone having some new and or undocumented way of generating text.