Stats on compact block reconstructions

The email provides a detailed analysis of transaction dynamics and node behavior over several months, highlighting the changes in transaction handling and block formation. Initially, the node's efficiency in processing transactions without requiring round trips significantly dropped from early July to late July and has since remained low. Specifically, the percentage of blocks needing no round trip decreased sharply from a range between 45–55% to just 2–15%. Concurrently, there was a notable increase in the median number of transactions per block, escalating from 2–3 to between 54 and 91.

A significant portion of missing transactions within this observation period pertained to Rune mint chains. The data indicates various reasons for these omissions, with 56% of transactions never being announced to the node before their respective blocks were formed. Another 37% involved conflicts where the node's first missing transaction conflicted with an existing version held by the node. Additionally, 7% were transactions that had been announced prior to block formation but hadn't yet entered the mempool.

The behavior of transaction announcements and acceptance also varies depending on the timing of block findings. Blocks found within 60 seconds of the previous one contained significantly more transactions (median of 454) than those found after a gap of three minutes or more, where the median number of transactions generally fell to zero. This suggests a quick succession of blocks favors higher transaction inclusion, particularly for those released in batches post-block confirmation.

Further insights reveal that the nature of transaction mining remained constant around mid-July, with no changes in the share of Rune transactions, chain depth, or the pattern of 24-transaction windows. However, off-chain dynamics such as the speed at which transaction windows and replacements reach mining pools compared to their distribution across the public network, alongside the volume of replacement traffic, have changed. Notably, these factors did not stem from any modifications in the node’s build, including PR 34628, as confirmed by consistent performance in OCEAN block reconstructions using the same binary setup from early July.

Overall, the data underscores a shift in transaction handling practices and network interactions affecting transaction propagation and block assembly, particularly impacting how quickly transactions are shared and incorporated into new blocks on the network.