How the JSON is read
The JSON parser in CSVUtils keeps each number’s original text, so the id 9007199254740993 in the example arrives in the CSV unchanged, and 1e2 stays 1e2 until you choose to convert it.
Records with different keys are merged into one header in first-seen order. Keys a record lacks become missing cells. Explicit nulls and absent keys both become missing, and the import notes say how many of each there were.
Duplicate keys inside one object (allowed by the grammar, undefined by the spec) are reported and the last value is used. Choose the nesting and array options on the source’s import settings, then export from the Export tab.
Questions
Can I paste JSON instead of opening a file?
Yes. Use “Paste data”; text starting with [ or { is read as JSON or JSON Lines.
What happens to arrays like ["office","light"]?
By default they are kept as compact JSON text in one cell. Choose “Spread to path[0], path[1]…” to give each element its own column.
Is the JSON evaluated as code?
No. It is parsed by a strict, depth-limited parser; nothing is evaluated.