The convenience is not worth the issues it can cause. There are equally convenient alternatives that don't cause any issues.
The main idea here is two-fold: doing this can cause problems, and there are other things you can do that don't cause any problems. First, the problems:
Someone might care about badges, and that is okay.
Stack Exchange is, by design, gamified via reputation points and badges. These things are a part of the system. Accumulating rep and badges is fun, and some users, especially new users care about them. And this is okay, it is what they are there for. We should not be dismissive or derogatory of the idea of caring about badges. To be clear, I very much doubt any high rep users are too much concerned about the badge thing. I am not going to complain about missing out on what might have been my 400th enlightened badge when I can get five more tomorrow. But a new user with zero silver badges might care.
Issues with the new user experience are well documented, both locally and on meta.se. And it is new users, users with no rep and no badges, that are most likely to express a meaningful level of concern for a missed badge, and I don't think it is right to tell them not to. Personally, I don't want to have this conversation with a new user:
Hey, so I know that it looks like you were supposed to get your first silver badge for this good answer, but what happened is a user with thousands of rep and hundreds of badges posted a non-answer and deleted it so they could come back and edit it later, which blocked your answer from being eligible.
This would be a lovely addition to the already challenging new user experience Stack Exchange presents (/S). If we can avoid creating circumstances where a new user might feel cheated out of something they have every reason to believe they earned, we should.
It can create issues when used to circumvent question closure.
The question that motivated this meta post was already at three close votes when the "answer" post was created and deleted. This allowed the answer to circumvent the closure of the question. Why is this a problem? The question's author returned and accepted the answer after the question was closed, and now has no reason at all to resolve the issues with the question.
See this meta: Allow questions to be saved as drafts prior to posting In particular:
If you start answering a particular question, but do not successfully submit, you will see your last saved answer draft the next time you visit that particular question page.
- We save drafts automatically for all new answers and new questions once every 45 seconds.
- Drafts are not supported on self-answer.
- Drafts are not supported on edits.
- You only get one draft for an answer and one for a question. (If you start a new post, the old draft is cleared.)
- Each site has its own draft storage, e.g. you can have multiple draft questions saved on multiple StackExchange sites
- Drafts will be automatically cleared after a week.
- Drafts work for anonymous users as well.
If you start working on an answer, the system saves your work every 45 seconds and keeps it there for an entire week. There is no reason to submit the incomplete answer because the system will save it for you. "Just don't submit the answer until you're done" has all the same convenience as submitting the incomplete answer, but causes none of the issues outlined earlier.
Alternatively, you can use the Formatting Sandbox on meta.se. Start an answer or edit an existing answer there and your work will be saved for time immemorial. It may get deleted in the distant future, but it will certainly be preserved for long enough to finish your answer and transfer it to the destination Q&A.
Finally, just use a notepad on desktop or mobile. This is what I usually do. If I want to move on and work on other stuff before I finish an answer, I'll transfer it to a .txt file on notepad before writing other posts. It's just as easy as creating an incomplete answer, and it has exactly zero potential for causing problems on the site.