Most of our list methods (search, discover, popular, etc...) are capped to the first 20,000 items for performance reasons. Elasticsearch struggles with this and so it is restricted in code. Newer versions of ES have improved this a lot but we're still on quite an old version.

The advice I have given people in the past is that if they need data on more than 20,000 items like this, it's probably better to bring the data in house and then you can run whatever queries/filters you want locally. Ie. by iterating over all of the available media ids.

and it also produces a lot of deleted movies

I don't understand this comment, can you elaborate? Do you mean that by iterating over the ids that you get a lot of 404's? If so, that's exactly why the daily ID exports exist. Grab a recent file and iterate over these ids instead, while there could be a few 404's (those deleted since the file was created) you're basically going to be 404 free.

Yes, 404. How to open json.gz file?

Not sure what platform you're on, best to just Google how to extract a .gz file. It's a gzipped file.