---
title: "Comment by Ahmet Özel on \"The Tester's Guide to RAG Evaluation, Part 1: Foundations\" | Hashnode"
description: "Fresh sessions are an important isolation boundary, but they may not reset application-level answer caches, workspace memories, or retrieval reuse outside the c"
url: "https://hashnode.com/posts/testers-guide-to-rag-evaluation-part-1-foundations/6ae4c1d35ceaf88286f34330/comment/6ae79116b1e56f7a74748858"
---

[Discussion](https://hashnode.com/posts/testers-guide-to-rag-evaluation-part-1-foundations/6ae4c1d35ceaf88286f34330)

Comment [The Tester's Guide to RAG Evaluation, Part 1: Foundations](https://hashnode.com/posts/testers-guide-to-rag-evaluation-part-1-foundations/6ae4c1d35ceaf88286f34330)

[Ahmet Özel](https://hashnode.com/@ahmetozel)

[Nov 1](https://hashnode.com/posts/testers-guide-to-rag-evaluation-part-1-foundations/6ae4c1d35ceaf88286f34330/comment/6ae79116b1e56f7a74748858)

Fresh sessions are an important isolation boundary, but they may not reset application-level answer caches, workspace memories, or retrieval reuse outside the conversation object. A new thread can therefore still inherit state that changes what the next case sees.

For the spike, I would run the same target case both alone and after a deliberately different predecessor, then compare the actual context records rather than requiring identical generated wording. Recording cache-hit flags and workspace/session identifiers in the trace would help explain any order dependence before it gets mistaken for a model-quality difference.

[Ashen Jayawardena](https://hashnode.com/@cinderloop)

[Nov 2](https://hashnode.com/posts/testers-guide-to-rag-evaluation-part-1-foundations/6ae4c1d35ceaf88286f34330/comment/6ae79116b1e56f7a74748858#reply-6ae838fefa2e89bfc5ee728d)

Thanks, this is a good catch. I'll update the article to cover this!

[Ahmet Özel](https://hashnode.com/@ahmetozel)

[Nov 6](https://hashnode.com/posts/testers-guide-to-rag-evaluation-part-1-foundations/6ae4c1d35ceaf88286f34330/comment/6ae79116b1e56f7a74748858#reply-6aedf792d00d08a4784e2354)

A compact example could make the update especially testable: run a target case alone, then after a predecessor that primes a different workspace fact. Compare the retrieved context and cache flags for the target in both runs.

If those inputs match but the generated wording changes, that is a different class of variation from inherited application state. Keeping those two outcomes separate would help readers avoid diagnosing every nondeterministic answer as session leakage, while still catching leakage that a fresh chat does not reset.
