Trying to fuzz my current install to understand more about the issue. Have switched the columns to utf8mb3_general_ci collation which changed the results slightly but not meaningfully.
Now looking at the yarpp options – I have changed the settings to only “consider with extra weight” for the title & “do not consider” for everything else. I can see that the results re-generate when I change this setting. Looking at a different post which is titled “Butterfly updates”, the related posts I get are “Bindweeds”, “Woodcock Wood”, “Tree planting”, etc – but if I use the WordPress search box I get lots of results with post titles including “The Big Butterfly Count”, “Butterfly management & conservation”, “Big Butterfly Count 2021”, “Big Butterfly count – 2020”, etc. So something is clearly going wrong with the matching. I guess I’m going to try and isolate a particular SQL query that’s producing these poor matches next.