Failure cases you should see

A scientific package should fail visibly when assumptions are unresolved.

Long missing interval

If a long gap remains after explicit resampling, fit_fpca() raises an error rather than imputing it.

Duplicate timestamps

from_long_dataframe() refuses duplicate times within a curve. Decide upstream whether they are duplicates, simultaneous binocular samples, or another data structure.

Mismatched layouts

The package cannot know that (0.8, 0.2) means “source card” in one stimulus and “decorative image” in another. Coordinate harmonization requires substantive design knowledge.

Constant channel + variance scaling

A constant functional dimension cannot be standardized to unit integrated variance and triggers an error.

Automatic registration

There is no automatic registration step. Define landmarks or explicitly call an elastic workflow.