DEFAULT_INSTANT_UNIT

The precision an Instant column is written with when no target Schema is supplied.

Microseconds, not nanoseconds, for range: an Arrow nanosecond timestamp is an int64 count of nanoseconds since the epoch, which only spans 1677–2262, while a microsecond one reaches roughly ±292 277 years around it. That still does not cover the whole StdlibInstant range — an instant beyond it is reported as ConvertingMismatch.ValueOutOfRange and refused rather than silently wrapping — but it covers every instant a file is likely to hold. It is also the precision pandas, Polars and PyArrow emit by default. The trade-off is that a sub-microsecond instant loses its last three digits on write; that is reported as ConvertingMismatch.PrecisionReduced, and an explicit target Schema with Timestamp(NANOSECOND, "UTC") avoids it.