(1) and (2) greatly depend on field. A lot of people are discouraged using the active voice in, for example, an academic thesis because it allows the nasty ques

(1) and (2) greatly depend on field. A lot of people are discouraged using the active voice in, for example, an academic thesis because it allows the nasty question of "So how much of this did you do?" and writing "I did this" comes off as a bit weird. In journal papers the use of the active voice is slowly improving. I still find it a bit odd that people use we for single author work, but that's probably just ingrained. To address point (3) a common route is to add an appendix or supplementary material. This is commonly done in deep learning for derivatives of novel functions so that others can implement the backward pass. Whether you can get it past a supervisor is another issue entirely! A blog post or accompanying website is an easy option. (4) is another tricky one. How many times have you read "Deep learning has revolutionised <field>" followed by the usual citing of Hinton, Le Cun, Krizhevsky et al.? The spiel is identical in every paper, how many ways can you describe this stuff after all? At some point you just have to assume that the reader is familiar with the background, but you still have to cite everyone to keep the reviewers happy. Again, if you need exposition, write a blog post about it. (This isn't just for deep learning - every field I've worked in suffers from the same copy-pasta introduction.) This also depends again on field anyway - some journals are much more approachable than others. Nature and Science, controversy aside, generally publish very readable papers.