ProbeLab reports that changing its Polkadot crawler substantially increased the number of reachable peers it observed. The finding concerns measurement coverage, not evidence that an equivalent number of new machines joined Polkadot.

In research published on 16 October, Dennis Trautwein attributes the earlier undercount to connection-attempt limits interacting with a short deadline. The reported average rose from 753 to 1,443 reachable peers per crawl; the share reached again in the next crawl increased from 65% to 96%.

What changed—and what NetNapz checked

The detailed experiment says that on 7 October the per-host file-descriptor dial limit rose from 160 to 2,048 across two hosts, giving 4,096 slots in total. A subsequent increase to 4,096 per host produced little further improvement. These are the researcher’s results; NetNapz has not rerun the crawls.

NetNapz inspected the tagged go-libp2p v0.50.0 dialer source. It defines ConcurrentFdDials as 160 and DefaultPerPeerRateLimit as eight. The comments distinguish concurrent outbound attempts consuming file descriptors from the limit applied per peer. This corroborates the cited defaults, not ProbeLab’s measured network counts.

An unresolved networking question

In a 19 October Polkadot forum post, Trautwein separately urged networking developers to examine stale listening-address records. He warned that connection deadlines can expire before a working address is tried. That is an attributed diagnosis and request for investigation, not a confirmed network-wide outage or an announced protocol fix.

NetNapz assessment

A changed observation method creates a break in a time series. Comparing the older and revised counts without that qualification risks turning better visibility into a misleading adoption headline. Reachable peer identities also do not, by themselves, measure independent operators, validator participation, transaction demand or economic decentralisation.

The practical test is reproducibility: independent collectors should disclose their client version, concurrency settings, deadlines and coverage checks before comparing results. Persistent connectivity problems would require separate operational evidence. Neither this crawler correction nor the forum warning establishes a DOT price effect, token demand or a change to consensus safety.

Sources and verification limits

• ProbeLab: 160 Dials, published 16 October 2026; experiment and measurements attributed to the research team.
• Versioned go-libp2p source and 19 October forum discussion linked above. Sources checked 20 October 2026.

Source-code inspection is not independent reproduction of network measurements. No live price or complete network inventory is asserted.