This is an unverified issue from reviewing existing code. But maybe we can find some examples where it affets results.
In the indexer we represent the Title field not as a single field, but as many fields, one per parsed term. We also apply a boost of 0.9 to each field.
According to this doc on lucene default similarity implementation:
"If the document has multiple fields with the same name, all their boosts are multiplied together: "
A probably unintended consequence is that our boost for title field will get exponentially smaller as the number of title terms increase.
Additionally we probably won't be able to search for a phrase like "JQuery UI" using quotes, since it's stored as two separate fields JQuery and UI, with no position relationship to each other.
Addressed by #671