Subsequent extra-info descriptors covering the same measurement periods change read/written bytes values
While sifting through extra-info descriptors with @juga we came across a weird thing in a bunch of read/written bytes lines. Take for example relay 85D13148908D1397278949A9850C6AEAEEDF6DA3 and the attached extra-info descriptors from 2024-08-12. Here are the relevant values from them:
| published | end of most recent interval | read-history | write-history |
|---|---|---|---|
| 2024-08-12 17:16:4783cc2a0e04ef97891d6f6ffd753da41c4094f9ff | 2024-08-12 15:23:37 (86400 s) | 156902400,126316544 | 67650560,69638144 |
| 2024-08-12 18:29:16e116c1e63e12cb09e2d2d63fa32ba3aa5276a274 | 2024-08-12 15:23:37 (86400 s) | 156902400,125970432 | 67305472,69292032 |
| 2024-08-12 19:53:34784aba7310681166015ec365a226c3ac0ad75437 | 2024-08-12 15:23:37 (86400 s) | 156902400,125970432 | 67305472,69292032 |
| 2024-08-12 19:53:41243711d9db41fd15183c1456ad8236686adbd803 | 2024-08-12 15:23:37 (86400 s) | 156902400,125970432 | 67305472,69292032 |
| 2024-08-12 20:01:011a4ce2c88fe7c59dc1c76abfeefd6093728b876a | 2024-08-12 15:23:37 (86400 s) | 156902400,125884416 | 67218432,69206016 |
Even though there have been 5 descriptors published during that timeframe the read/write-history is related to the same period: the latest one ending 2024-08-12 15:23:37 and (implicitly) the previous one ending at 2024-08-11 15:23:37. And, given that all five descriptors got published later than those periods there should not be any changes to the respective read-/write-history values. Yet, we see that they get smaller not only for the period covered last but even as well for the previous one. They don't seem to get larger (again), though.
We have seen other relays with similar behavior which seems to indicate that we have a bug rather than someone manipulating those values for whatever reason.