Thanks for clarifying @Dana_Pavaday. It would be useful to have a rough idea of size of the field.
I did find this old thread and the blog post mentioned in that thread that talks about the underlying Lucene limit that I wonder if you might be hitting:
And even if you do raise the Elasticsearch limit, you cannot exceed the hard Lucene limit of 32k for a single token, and Elasticsearch certainly logs some messages that exceed this.
I would recommend having a look at setting your ignore_above limit in line with the note at the bottom of the documentation and seeing if that helps when the document is reingested.