What is written
The hash of the account's key, its two balances, and the order names it has already seen.
What happens to the fills
Reading an account back builds it with an empty run of fills, whatever the file held — and the file holds none, since writing never names them.
A fill is appended to that run as it is applied, and the run is the only record of one. So the balances outlast a restart and the account of how they came to be does not.
What a reader sees afterwards
Asking a restarted ledger for an account's fills gives nothing back, for an account whose balances show plainly that it has traded.
The numbering carries on
The number the next trade will take is written and read back, so a fill made after a restart is numbered past the fills that are gone.
A reader of the numbers therefore sees a run beginning partway up, with nothing to say where the missing part went.
What else survives
The order names an account has seen. A repeated order name is turned away after a restart as before it, so the one guard built on remembering survives while the record it would be checked against does not.
Related
Source evidence
---
row: P64-06
baseline: 49aa659
created: 2026-11-08 19:39 UTC
modified: 2026-11-08 19:39 UTC
evidence:
- claim: "Writing an account out names the hash of its key, its two balances and the order names it has seen, and names its fills nowhere"
source: "`ledger.py` 754-759"
- claim: "Reading an account back builds it with an empty run of fills, whatever the file held"
source: "`ledger.py` 771"
- claim: "A fill is appended to that run as it is applied, so the run is the only record of one"
source: "`ledger.py` 446, read out at 613"
- claim: "The numbering a trade is given carries on from the file, so a fill after a restart is numbered past the ones that are gone"
source: "`ledger.py` 251, the number taken at 435 and 562 and carried on at 447 and 571, written out at 625"
- claim: "The order names an account has seen do survive, and the test that turns a repeated one away reads them"
source: "`ledger.py` 758, read back at 764 into 772, read by the test at 534"
---