Chaharshanbeh Suri is the fire festival held on the eve of the last Wednesday before Nowruz. That is its whole definition: it is the last Tuesday night of the old year, and everything about it — the bonfires, the leaping, the clearing out — points forward at a new year that has not happened yet.
Our records placed it, in 2026, on two different dates in two provinces of the same country:
| one province | 20 March |
| Tehran | 25 March |
Nowruz 2026 falls on 21 March, so the festival belongs around the 17th or 18th. The first date is wrong by a couple of days. The second is four days after Nowruz, which a festival defined as preceding Nowruz cannot be under any reading.
The dates were not the defect
It is tempting to correct two dates and move on. That would have left the fault untouched, because both had been modelled as fixed calendar dates.
A fixed date cannot express “the eve of the last Wednesday before Nowruz”. It cannot express it approximately, or usually, or in most years. The rule type has no vocabulary for the relationship, so whoever filled the record had no choice but to write down a date that was right in whatever year they were looking at, and let it drift from then on.
Two provinces did that in different years. Hence five days.
This is the more useful way to read a wrong date: not as a typo, but as evidence about what the model could not say. A date that is wrong by a couple of days suggests transcription. A date that is wrong in a way the definition forbids — after the thing it precedes — suggests the definition was never captured at all.
How it was found
Nobody went looking. A structural check fired on the province’s file for an unrelated reason: the file had been emptied around this record by earlier work, leaving one observance whose day was already closed nationally, which is a filing question rather than a calendar one.
The real problem was sitting underneath the filing question, and it was visible only because the corpus now held both provinces’ answers in one place and they disagreed. The same principle as one full moon resolving three ways: a disagreement between your own records is evidence, and it costs nothing to collect.
The filing question was the smaller half of what the check reported. It is frequently the smaller half.