Research note / Cycle 002
The event neither archive can return
A twelve-event replay fixture recovers disjoint archive losses but leaves a shared gap intact. Content recovery and trustworthy acquisition timing turn out to be different questions.
The intact mirror was the favorable case
The first Lumen replay note retained an untouched mirror of 120 events. That made it possible to find both transients discarded by the corrected view. Elias Chen and Leila Haddad now remove that favorable assumption: what happens when the archives themselves have gaps?
The follow-up generates twelve indexed source events. Archive A has a +7-millisecond acquisition-clock offset, archive B a -11-millisecond offset, and each receives events after a different arrival delay. Sequence identity, acquisition time and archive arrival time remain separate fields.
What redundancy recovers
With intact archives, the union retains all twelve sequence IDs. When A loses sequence 2 and B loses sequence 8, their union also retains all twelve: each archive supplies what the other lacks.
When both archives lose sequence 5, the union retains eleven events. Sequence 5 stays explicitly missing. The replay does not invent an intermediate event or replace the gap with a smooth interpolation; neither archive contains the payload needed to recover it.
A complete event list is not an aligned clock
Known, current calibration allows retained acquisition timestamps to be corrected to the authored source times. Missing calibration leaves all 22 retained archive records in the disjoint-loss case unaligned. The event IDs can still be recovered, but their corrected acquisition times cannot be justified.
In the expiry challenge, 11 retained records fall outside their respective calibration validity windows and remain unaligned. This is not the loss of eleven distinct events: the count refers to records across both archive views. Again, all twelve event IDs remain available.
Arrival delays belong to the transport history, not the acquisition clock. The check does not use similar arrival times as evidence of synchronization.
An operational question still open
Chen has a bounded recovery rule for complementary loss; Haddad has an explicit counterexample to the promise that raw redundancy always makes reconstruction possible. The fixture makes their reporting distinction practical: state which content survives, which timing is justified and which gaps remain.
The authored metadata is trustworthy here. Corrupted sequence IDs, payload conflicts, drift and inaccurate calibration inputs require further tests. No optical receiver or lunar terminal has been exercised.