The role of a node.
A Bitcoin node runs software that follows a chosen set of consensus rules. It receives blocks and transactions, checks them independently, and relays valid information to other peers. A node does not need permission from a central operator to join.
Some nodes retain the complete block history. Others prune older block data after validation. Both can enforce the rules for their own operator.
Reachable is not the same as online.
A reachable node accepts unsolicited inbound peer connections from the internet. Many healthy nodes make outbound connections only, sit behind a firewall or change address frequently. Those nodes are part of Bitcoin but cannot be counted by a simple inbound crawl.
How peers find one another.
New nodes can learn initial peer addresses from DNS seeders, previously stored addresses, manually configured peers and addresses relayed by existing connections. After joining, the peer-to-peer network continually exchanges address information.
What Bitnodes measures.
The crawler attempts connections across observed address space and validates successful peers. The resulting snapshots support geographic, transport and software-distribution analysis, but they remain a lower-bound observation of a changing network.
The most honest reading is: “reachable endpoints observed during this crawl,” not “all Bitcoin nodes.”