v0.66.16

Managed workspaces recover on replacement hosts

Restoring durable receipts now reserves their operation identities before new commands run. A replacement host can continue accepted work without colliding with its recovered history. Release archives are checked against this recovery case before their checksums are published.

Native commands retain a stable source view while peers publish changes. Perl in-place editing and directory moves preserve final ownership checks, and failed commands retain their diagnostics in rejected receipts. Claim waits can recover a missed grant notification with an authoritative check at their deadline.

The managed adapter now uses a Rust coordinator and persistent worker connections. Its public socket protocol has changed: upgrade the adapter and connection helper together and build the shipped Rust runtime on Linux before starting it. Existing application SDK calls remain compatible. The adapter remains opt-in; installing these packages does not enable it or deploy server-side claim changes. See the adapter guide for setup and runtime limits.