Evidence for learning behaviors in plants is to my knowledge very scarce. [...] a sunflower who's stem grows on the shadier side and reset during the night in order to "track the sun".

The learning behaviour is actually much more nuanced than that. This is one of the main papers on it (the Youtube video I linked earlier puts it in a more digestable format, plus provides some extra context from the main researcher's talks and other papers).

Basically they took a plant which has a reaction to close its leaves when disturbed and repeatedly dropped its container a small distance, which caused the plant to close its leaves. After only a small number of repeated drops, the plants started reopening their leaves more quickly, and eventually stopped closing them entirely. However, the plants would still close their leaves as normal when a different stimulus of gently shaking their container was presented. They were also capable of remembering the learned behaviour of not closing when dropped for months at a time, and their capacity for learning depended on environmental conditions -- when the plants were grown with less light, they learned more quickly.

In other words, it's much more than a reflex. I would like to note that the "it's just a reflex" argument doesn't have a very good track record in general!

The same researcher, Stefano Mancuso, has many other fascinating plant intelligence experiments and is one of the pioneers of the field of plant neurobiology. The pea plant sensing experiment I mentioned earlier is also his work. It is still an emerging field, but the evidence is quite impressive. It also just makes sense when you think about it -- there has been so much evolutionary benefit for sensing and learning behaviours in all other life down to the microscopic level that it would in fact be more surprising if similar behaviour didn't exist in plants.

I don't think anyone, including the authors, is honestly claiming it as incontrovertible evidence of plant sentience. Personally I think the main takeaway from it should be that learning, sensing etc. aren't nearly as complicated of a behaviour as we used to think they are, and can likely occur without consciousness or sentience as we humans experience them.

It is somewhat arbitrary very the exact line is, but I don't think there is much debate to be had that killing animals is unethical. There are very few people who wouldn't react with horror at the idea of killing kittens and milking the mom dry but somehow for cows this is acceptable?

Yeah like I said, the ethical argument is easy for complex mammals. It's simple animals where it's hard. I think basically no one, including most vegans, would find any ethical issues with killing a parasitic worm, for example.

I don't think it is possible to create a clear and objective definition of intelligence or sentience that excludes all plants (and mushrooms!) without excluding e.g. tardigrades, honeybees, silkworms, mealworms, snails, or indeed lobsters. We simply lack the metrics on which to base it. And I believe if we did have those metrics, we'd find that on them many plants indeed "outrank" many animals.

Even beyond the purely philosophical argument of defining a stronger foundation for veganism besides "is it in the kingdom Animalia", this has actual practical relevance. Certain micronutrients that cannot be gained (or easily gained) from plants would be available in simple animals, which could make them a valuable food source in a world that doesn't eat chickens, pigs or cows. Simple animals can also be valuable for bioengineering. Conventional non-food uses like silk production are also pretty nice.