polishing the beginning of the paper and leaving the end close to a mess.
awkward priorities (e.g.: the goal is to apply my method to save the planet; instead: the goal is to save the planet using my method). More generally, say WHAT you want to do, and then HOW.
mixing the details of the experiments with the algorithm/method. Explain the method; then you make the list of the parameters (tabular); then the goals of experiments; then the results and their interpretation.
have a mini-survey as state of the art (not just a name-dropper performance). The reviewer is always very happy to learn something.
when you plan to investigate a neighbour domain, find someone in your environment who is well versed in this domain, and share the paper writing/authoring. There is no way to invent the jargon; there is no way a paper can be accepted without the proper jargon.