Measuring minrelaytxfee across the Bitcoin network

The peculiar case of the 9170997 entries observed in Bitcoin nodes has sparked a detailed examination into the mechanics behind Bitcoin Core's handling of minfeefilter values. An investigation into these anomalies revealed that Bitcoin Core intentionally adjusts feerate filters to prevent potential node fingerprinting across networks. This adjustment is facilitated through the FeeFilterRounder mechanism within the Bitcoin Core source code, which specifically targets the optimization of feerate filters by placing them into predefined bins. The FeeFilterRounder holds a maximum filter feerate (MAX_FILTER_FEERATE) of 1e7 (10,000,000), a value that intriguingly rounds to 9170997. This rounding process is not arbitrary but is a calculated measure as reflected in both the code and associated tests found on GitHub, demonstrating how MAX_MONEY translates to the same 9170997 figure under certain testing conditions.

Further insights into this matter are provided by examining the behavior of nodes during the Initial Block Download (IBD) phase. It is noted that nodes set a feefilter of MAX_MONEY when in IBD, leading to the appearance of the 9170997 feefilter value. This observation suggests that the nodes displaying this specific feefilter value were likely in the midst of IBD at the time of connection. This hypothesis is supported by log examinations showing a recurring pattern where connected peers initially communicate a feefilter of 9170997, which is then reduced to 1000 after a brief period. Such behavior underscores the dynamic nature of feefilters within Bitcoin's network protocol and illustrates the intricate balance between privacy, security, and efficiency achieved through these mechanisms. The aforementioned links to the Bitcoin GitHub repository, such as the FeeFilterRounder and tests related to policy fees and net processing code (policy fee tests, net processing), provide concrete examples and further elucidate the technical rationale behind these observations.