monerosim - v0.2.0 - Turnover Tartare

Network topology realism + peer turnover modeling. This release makes monerosim's simulated network behave like mainnet along the axes that matter for topology research — connection-duration distribution and the long-lived-connection tail — at a realistic reachable fraction.

Highlights since v0.1.0

• --reachable knob. Models mainnet's NAT-unreachable majority (~15% reachable / ~85% unreachable). Per-agent --hide-my-port, override semantics, seeds + miners always reachable.
• Peer turnover. Relays and users cycle offline/online mid-run, reproducing real node churn. native_preemption: true fixes an LMDB DB-resize livelock on turnover restart.
• Turnover-robust Rucknium parser. Analysis survives restarted daemon logs (stock xmrpeers crashes on them).
• max-connections-per-ip fix. Floored at 4 to stop a P2P reconnect loop (back-ping refusal); surfaced as a documented, overridable daemon default.
• Network topology realism study. Reachability + turnover match mainnet on connection-duration median (~23-29 min) and the >6h tail at realistic 15% reachability. See docs/20260620_network_topology_study.md.
• Rucknium review response v2. Verified mechanisms, replication, corrected scope.

Result: at 15% reachable with 1h/1h turnover, connection-duration median ~28.75 min (mainnet ~23) and >6h connections at the supernode floor — mainnet-like on both metrics simultaneously.

Full changelog: v0.1.0...v0.2.0

Pairs with: shadowformonero v0.1.0 (the last pre-upstream-sync fork; pinned via SHADOWFORMONERO_REF in this tag's setup.sh — later releases record the pairing in shadowformonero.pin).