The crawler resolves a proxy for every request with ProxyConfiguration.new_proxy_info(session_id, request, proxy_tier) (_basic_crawler.py _get_proxy_info). Session stickiness lives in ProxyConfiguration._used_proxy_urls, an in-memory dict keyed by session ID, so it is lost on migration even when the session pool is persisted.

JS v4 assigns the proxy once, when a session is created, and stores it in the session.

Proposed solution

• Add Session.proxy_info, set by the session pool when it creates a session and persisted in SessionModel. Crawlers read the proxy from the session.
• ProxyConfiguration.new_proxy_info() / new_url() take no session ID or request. A restored session passes its previous ProxyInfo back so the configuration can reuse it.
• new_url_function takes no arguments; remove ProxyInfo.session_id and _used_proxy_urls.
• Warn when a crawler gets both a custom session_pool and a proxy_configuration, since the pool decides the proxy.

The protocol extraction itself is #2137. Removing proxy tiers is a separate follow-up issue.

JS references

• Implementation: feat: store ProxyInfo inside Session instances crawlee#3199 (closes Remove the notion of session from ProxyConfiguration crawlee#3198)
• request option dropped: refactor!: drop the unused request option from ProxyConfiguration crawlee#4153
• Restored ProxyInfo passed to newProxyInfo: fix: pass restored ProxyInfo to newProxyInfo crawlee#4198
• IProxyConfiguration: Interface isolation — IProxyConfiguration crawlee#3883, refactor: Introduce IProxyConfiguration crawlee#3892

This will be a breaking change.

✍️ Drafted by Claude Code