This one is unnecessary, in the new schema, attribute reads do not need to be deferred, but can happen instantly without performance penalty, once preloadAttributes has been filled.
If the streaming API is used on files, parsing the whole Series up front is unnecessary work. Our frontend does not yet allow to distinguish whether parsing the whole series will be necessary since parsing happens upon construction time of Series, but the classical and the streaming API are both activated afterwards from the created Series object. Hence, improving this requires refactoring in our user-facing API. Ideas: (1) Delayed lazy parsing of iterations upon accessing (would bring other benefits also). (2) Introduce a restricted class StreamingSeries.