TL;DR: I don't like the existing code bowling questions, but I'm not sure there is no hope for it. I'm not suggesting a course of action in this post, but just laying out some arguments for either side, so people can form their own (informed) opinions.
I don't agree with the parallels you're drawing between code-trolling and code-bowling. There are a lot of differences between the two. In particular writing long and unmaintainable code was just one way to approach a code trolling challenge. But code trolling wasn't about that. Code trolling was a lot more general (which was one of its main problems). Hence, I don't think you can form a valid argument here that is based on comparison with code trolling.
I do agree with code-bowling being problematic though (well, of course I do, your post quotes my own comment :D). However, I have to disagree a bit with you here as well. In particular, I don't think code bowling is lacking an objective winning criterion. Because "largest code size in bytes/characters" is equally objective as any code golf. The problem is coming up with loophole-free rules that can be enforced objectively.
I'm not saying it's impossible, but it's very hard. Statements like "no line must be useless" are not objectively enforcible. Not reusing functions is objective, but it leaves loopholes like just putting in an infinitely long literal or variable name. And I'm not even sure it would be a useful rule if it worked. You have to work really hard to get that down, and for the challenge to be fun as well, it should actually be hard to write a valid submission longer than 100 characters or so, because I don't want to slave away at writing 10k lines of code just to be competitive in the challenge.
Okay, so what I did was the same I did for code-trolling: I looked at every single code-bowling question that has been asked. I left out the two ones that are closed anyway, so it's down to 10:
- The very first code bowling is in fact really just code trolling with the restriction that the code has to work. This may have been why you drew the parallel between the two. It obviously lacks both objective rules and objective winning criterion.
- There are a few that are tagged both code-bowling and popularity-contest which basically amounts to code trolling as well: this, this as well as one of the closed ones.
- There are five which basically end in ridiculous quasi-infinite size submissions. They either ignore the fact that something like that would be possible, or tried to rule it out with things like "no redundant code" which you can always work around. Or if you start arguing that something is redundant you can very quickly extend the argument to any detour at all being redundant, which defeats the point as well.
- There is this question where I can't gauge if the challenge itself is actually impossible or - if not - if it would be trivial to extend to an infinitely long submission. But it's more of a code challenge anyway then actual code bowling, because the scoring is quite elaborate and just happens to incentivise convoluted code.
- There is this question which avoids the problem by putting a hard limit on the code size. I think this is actually a very decent idea, and probably the only half-reasonable instance of code bowling so far. Unfortunately, this particular question has another flaw, which is the not objectively measurable readability of the ASCII art. But that's beside the point here.
In general, I have to say, code bowling has been pretty crappy so far. But as opposed to code trolling, it doesn't seem to be something inherent to code bowling as a scoring mechanism. That last question shows (to me, at least) that with a bit of careful thought, a few good code bowling questions might be possible, and people just haven't been trying hard enough. At the same time there is an appreciable probability, that it's just too damn hard to come up with decent rules for such a question.
I think most folks writing a code bowling challenge just think "oh, let them write TONS of code, that would be fun, yay code bowling." Maybe we could change that vibe by rewriting the tag summary to something like "The goal of code bowling is to maximise the size of your code length subject to some constraint." And people would also have to make sure that the constraint has to be very rigid, because I don't think anyone wants to write (and much less read) those thousands of characters long submissions and just stop at an arbitrary point because they can't be bothered any more. You should have to work as hard for each single character in code bowling as you do in code golf. If you can come up with rules like that, I'm all for code bowling. As long as you don't I won't really grow fond of it.
With that in mind, I don't really know what course of action to suggest here. In two years, only one question managed to get anywhere near being a good code bowling challenge (in my opinion). So it might well be possible, we just need to try harder and keep looking. At the same time, I'm not really sure we will find the holy grail of code bowling any time soon, so maybe this concept doesn't warrant it's own tag (which just seems to attract badly thought-through challenges) and those new concepts should be given their own chance as a simple code-challenge. I don't think we need to be as harsh with this as with code trolling, but I wouldn't mind cleaning up most of the existing code bowling either, because currently it's not really doing any good.
So as I said at the beginning, I'm not actually voicing a clear opinion on what to do here, I just wanted to collect the above information here, so people can draw their own (informed) conclusions.
Long-ish PS: After thinking about it for a while, my own recent challenge is a bit of a code bowling (disguised as a code challenge). I asked people to print as many digits of the golden ratio on a strict character budget per digit. This is not much different from asking for the longest code that prints a sufficient number of digits, which depends on the code length. I had a really hard time closing loopholes to prevent infinite scores, which made the rules ridiculously complicated, and I wouldn't want every code bowling challenge to have to be that convoluted. However, I think I did manage to create a situation where - very soon after they started - participants (except the accepted answer) actually had to work hard for every single character, just like you do in code golf. But as you can see, the accepted answer, even though not scoring infinity, side-stepped all that and did find a loophole in the end. So while I wasn't aware of it, I've also fallen victim to the perils of posing a code bowling type question.