[abhs21]

RadLab diagnostics could reject valid nanosecond timestamps on Python 3.9 or merge distinct timestamps into false duplicates on Python 3.12. This PR preserves exact timestamp identity throughout the checks and adds an offline tool for reviewing the original readings, neighboring samples and source evidence.

The review packet retains all 58 differing-rate groups across 18 snapshots, with source rows, hashes, queries and saved notes. Row labels explicitly count the CSV header as row 1. Export all notes includes groups hidden by filters and preserves original rate strings and units. Hitaeshi Sehgal’s verified neighboring-sample comparison is credited; closeness to a trend does not establish a physically correct reading.

The sampling study includes every prescribed mask at 0/10/20-minute offsets and synthetic controls. All 28 prespecified May/June queries returned header-only CSVs; their bytes, hashes and exclusions are preserved without replacement dates. The numerical masking results still concern only two original eligible series and do not bound missing radiation dose.

A separate interval-accounting demonstration responds to the variable-integration question. Six analytic synthetic cases cover constant, linear and peaked signals, partial bins, gaps/conflicting intervals, independent gain scaling and known clock offsets. It accepts explicit wall-time interval means and durations, conserves complete-interval dose, reports coverage and partial-bin bounds, and rejects invalid/overlapping input. No native durations are inferred from export cadence. A real DosTel2/SPACEDOS02 comparison awaits documented interval and processing semantics.

Validation:

• All 67 tests pass on Python 3.9.6 and 3.12.14, including actual JavaScript export serialization.
• Browser save/reload, filtering and downloaded export verified with two instrument groups.
• Synthetic JSON and report reproduce byte for byte across both versions.
• Original source snapshots, source-packet JSON and baseline results remain unchanged.

Sampling report · Interval-accounting report · Project brief

The project brief requests a separate subgroup with Abhi as lead. Formal recognition and instrument interpretation remain pending.

Continues #2 after renaming the working branch. The code and commit history are unchanged.