State Without a Landlord: An Architecture Proposal for Peer-to-Peer Replication of Durable Workflow State

"State Without a Landlord: An Architecture Proposal for Peer-to-Peer Replication of Durable Workflow State" (arXiv:2609.17645v1) is a proposal by Luca Maraschi and Matteo Collina that relocates durable workflow journals from centralized providers to authenticated, append-only logs replicated among workflow parties.

Key architectural components include: P2P Replication: Uses Hypercores for signed, single-writer epoch chains and BitTorrent v2 for large payload distribution. Deterministic Replay: State is reconstructed by any authorized peer holding the journal and code, removing reliance on a central "landlord" orchestrator. Security Model: Employs quorum-finalized closure certificates and fork proofs to handle executor succession, safety eviction, and equivocation detection under crash-only or rational failure models. Design Principles: Enforces that the journal is the source of truth, coordination is separated from distribution, and trust needs are met by cryptographic evidence rather than implicit authority.

Submitted on September 15, 2026, this 51-page paper argues that durable execution journals are isomorphic to peer-to-peer logs, enabling state portability and independently verifiable custody without changing workflow source code.

State Without a Landlord reframes durable execution by focusing on the object that actually makes a workflow durable: the journal of executed steps. In conventional durable-execution systems, a long-running function is checkpointed by recording completed actions or effects, and later recovery is achieved by deterministically replaying that journal. Because the journal is the authoritative state, the provider that hosts the run also effectively becomes the custodian of the workflow’s state. The paper examines what changes if that journal is no longer a provider-bound artifact but an autonomous, peer-to-peer replicated one.

The key contribution is an architecture proposal for moving the workflow journal out of a single landlord-style provider and into a replicated P2P substrate. The central design problem is to preserve the semantic guarantees of durable execution—deterministic replay, ordering, durability, recoverability, and integrity—while changing who stores, serves, and controls the journal. That requires treating the journal as a verifiable, portable state object that can be replicated across peers, reconstructed after failure, and replayed by an executor even when the original hosting provider is unavailable or untrusted. The proposal therefore shifts the trust boundary from “trust the execution provider’s database” to “trust the replicated journal protocol and the cryptographic or consensus mechanisms that secure it.”

The work matters because durable workflows often represent long-lived, high-value processes, yet current deployments make that state hostage to a single provider’s infrastructure, pricing, availability, and governance. A P2P model for durable workflow state could reduce vendor lock-in, improve resilience, and allow workflows to outlive any individual execution environment. More broadly, the paper connects durable execution with decentralized storage and state-replication ideas, suggesting a path toward workflow systems whose state is portable, auditable, and not dependent on a single “landlord” for survival.