Twitter imports adds extra N to each word on many tweets, what could cause this?

That’s weird, I’ve never seen it doing that, and it’s not doing it on my site as far as I can tell.

Can you explain exactly where the “N” is being added? And maybe post some of the raw import data and/or a link to a post where you’re seeing it? You can get the raw import data by going to the “Edit Post” page for the post in question, within wp-admin. Then make sure the “Custom Fields” section is showing, and there should be one called “raw_import_data”. Copy-paste the entire contents of that into a reply here, and maybe I can see something weird going on there.

Yes, it adds “N” letter when the text goes to the line in your tweet.
Must detect paragraph and adds this.

example from real tweet of my timeline :

Final Fantasy VII sur ma cha?ne #twitch à 22h30
Aéris a quitté le groupe en pleine nuit !
https://t.co/QQOyR74IHZ

here, notice the “n” added whenever I go to the line in my original tweet :

Final Fantasy VII sur ma cha?ne #twitch à 22h30nAéris a quitté le groupe en pleine nuit !nhttps://t.co/QQOyR74IHZ

In raw data import “\n” appears in these spaces.
And “\n” means “new line” in php…

If you need to inspect, here is partial raw data for this same tweet :

{"created_at":"Wed Feb 24 21:07:33 +0000 2016","id":7.0260077749647e+17,"id_str":"702600777496469506","text":"Final Fantasy VII sur ma cha\u00eene #twitch \u00e0 22h30\nA\u00e9ris a quitt\u00e9 le groupe en pleine nuit !\nhttps:\/\/t.co\/QQOyR74IHZ https:\/\/t.co\/dCrr8SXiGZ","entities":{"hashtags":

Thanks for your work Beau, hope you can fix this soon ??

Not an expert but the lead might be here :

// Clean up post content for insertion
$post_content = esc_sql( html_entity_decode( trim( $post_content ) ) );

Are you by any chance using a Windows/IIS server to host your site? I know they are known to have trouble parsing different linebreaks sometimes.

The raw data there looks correct, I thought it might have been double-escaping those new-lines for some reason, but it’s not.

I’ll have to try it out and see if it’s a universal problem, or it might potentially be something weird to do with decoding the JSON/parsing entities, as you’ve suggested above.