docs(task_history): the acceleration refresh task is acceleration_refresh Eighteen deployment guides (and their 2.0.x-2.3.x snapshots) told readers to correlate object reads and accelerator spans with a task named accelerated_table_refresh. No release has ever recorded that name: the refresh span is declared as "acceleration_refresh" in the accelerated-table refresh task, and the task column is the span name unless a task_override label is present. The name a reader pastes into a task_history query has to match exactly, so a fabricated one returns no rows. cli/reference/trace.md already carried the correct name on these same snapshots.

docs(http): http_headers and OAuth2 params do not work on structured HTTP file datasets A structured HTTP file dataset (csv, parquet, arrow, avro, jsonl, ...) is served by the object-store listing path. HttpListingConnector forwards only port, username, password and a client_timeout fragment to the object store, so http_headers never reaches the request — yet the 'Using Custom Headers' section demonstrated it on a .csv URL, and both OAuth2 sections told readers to use http_headers on that path instead. Since spiceai/spiceai#12321 setting any OAuth2 parameter on such a dataset is rejected at registration rather than silently dropped, and ignored http_headers are logged as a warning. - Custom Headers examples now target a dynamic JSON API endpoint, with the listing-path exclusion stated up front and Basic auth given as the working alternative for a structured file download. - OAuth2 applicability notes and Limitations bullets corrected; the vNext text states the rejection, the versioned snapshots state that the params are not applied (which is what those releases do). - http_headers and auth_token_url parameter rows carry the applicability, matching the ParameterSpec descriptions. The structured-format list is scoped per release: vortex is structured only in 2.1.x and vNext, and 1.9.x-1.11.x recognize just csv/tsv/parquet/arrow/avro.