Issue with word_delimiter and minimum_should_match about elasticsuite

[rbayet]

TL;DR version: the multi_match does not handle correctly a minimum_should_match at 100% with queries that, once analyzed, might produce several tokens at the same place.
Which will be the case for "ProDuct" which will generate "product" (pos=0), "pro" (pos=0) and "duct" (pos=1).

Solution: Switching "split_on_case_change" to "false" in the definition of the "world_delimiter" filter in etc/elasticsuite_analysis.xml (see below) should do the trick, while losing the (dubious) benefit of being able to handle the "StarWars/starwars/star-wars" case.

        <filter name="word_delimiter" type="word_delimiter" language="default">
            <generate_word_parts>true</generate_word_parts>
            <catenate_words>true</catenate_words>
            <catenate_numbers>true</catenate_numbers>
            <catenate_all>true</catenate_all>
            <split_on_case_change>true</split_on_case_change> <== change this to false, clear caches and reindex
            <split_on_numerics>true</split_on_numerics>
            <preserve_original>true</preserve_original>
        </filter>

Enabling the experimental Search Relevance settings "Spellchecking Configuration > Term Vectors Configuration

[Experimental] Use all tokens from term vectors" is not guaranteed to provide any help, on the contrary in such case.