Thirteen restored tabs do not necessarily mean the same work survived. Account assignments or drafts can differ while the count stays unchanged. The recent retention check uses quantity as a starting point, then examines ownership, actual authentication and manuscript content through different evidence.
Match identities and relationships first
The added actual 0.1.211 check compares tab ID, profile ID, project ID, URL and pinned state before and after. Projects separately compare ID, name, default-account setting and tab count. Zero new tabs becomes meaningful alongside those retained relationships.
All three project defaults were unselected at the start. Preserving an empty default is not successful configuration of a new default account. Per-tab assignments and project defaults are different records too. The ledger should explain which relationships remained and which fields began empty rather than accumulate an impressive collection of counts.
Stored account labels and actual authentication differ
Matching saved display names and choices does not prove a valid website session. The actual check observes the specified publishing screen and authenticated screens under two Google profiles without logging in again. A separate unassigned tab retains its login button, checking that it did not inherit another account.
That concerns the four used tab profiles and named site paths, not every authentication across eight stored platforms. It did not read or duplicate cookies, passwords or authentication databases. Actual names remain private; public prose retains status and comparison method. Explaining session continuity requires no disclosure of the person's identity.
Wait and read a late account indicator
Immediately after the second restart, Gemini showed only a generic Google icon and strict account-fingerprint comparison failed. The report did not call that success; it waited by reading the same tab. After loading, the original account indicator matched, while the initial observation remained retained.
The change was a QA reading wait, not product code. Describing it as a storage-defect fix would distort the production record. Declaring an account change from the first generic icon could also be wrong. Dynamic-screen comparison needs a criterion for when an observation can support a stable-state claim.
Do not confuse a new-post URL with a saved manuscript
A Naver new-post URL may reopen as an empty editor or a restoration notice. Verification loads the exact-title existing draft and compares content SHA for two manuscripts. It does not claim that the generic new-post URL became a particular document address.
The record retains the step preserving current content before choosing the saved draft. Visible text alone is weaker than a same-content comparison. This still does not guarantee retention through provider expiry or power loss. A specific fingerprint and restoration route define the checked claim so another operator can interpret what survived and how it was observed.
Separate unchanged code from intended observation changes
The added check compares seventy-eight installed payload files and 208 public-verified runtime source files unchanged. No product defect was reproduced, so it made no app change or new installation. The earlier 795 regressions are not added to this actual-check count. Additional observation under the same version differs from a new product change.
This content work also reads saved reports without changing the original app or account. The shared comparator uses aliases, digests and exact planned differences while actual authentication, installation and restore remain false. The useful maker lesson is the criterion behind sameness. Explaining compared state and checked dynamic differences connects preservation to the person's actual work.