I have to make some assumptions here. Firstly this is not a one off. Secondly that you have noticed a trend. Thirdly it is a significant trend. Fourthly you have only looked a small sample and alerting us.
How many samples have you looked at and how many match your observations?
Furthermore this could only be done with Betfair as the oncourse flucs are a result of human manipulation. Similarly it could not be done on tote prices as the weight of money on any horse affects all prices.

The Betfair flucs are matching of a lay bet and back bet, and is a matter of accepting what is available at the time. Each and every matched bet creates a fluc. Hence this would need to be automated as some races must generate thousands of flucs. The Cup would most likely be the benchmark, so how many flucs occur in that event?