Yeah, that's daft. Most journals I use now have a "your paper your way" policy for initial submissions and only require conforming to format requirements when it's near acceptance. I'm quite happy to use their latex template at that stage particularly if the journal is open access and not charging exorbitant fees.

> Which means quite a lot of menial work to redo the article each time you apply, get refused, and want to send to another journal.

What kind of menial work do you mean? Isn't that just a matter of changing the class of your document? Why would you need to change the text of the article itself?

I'm not GP but some journals have formatting requirements such as specifying which sections (e.g. introduction, background, methods, results, discussion) you must have and in what order. Also relative size of these e.g. how much background is expected? Word/page length limits both for the abstract and complete article. Nit picks like whether you refer to Figure 1 or fig. 1 in the text (and whether it's capitalized if not at sentence start). Usually if you reformat a bibliography to a different referencing style, even automatically, it needs checking as the database fields will have been used in different ways and often errors creep in. Different wording for competing interests and author contributions statements and maybe a different position for the former (in the main manuscript / a separate file). The list goes on...

Not to mention often the latex style (or word template) given often directly contradicts the author instructions page on the website.