Version 2.0, 16 September 2026. Frame: all 19 consecutive IBDB season-opening lists, 2000–01 through 2018–19. Frozen before extracting production outcomes. 782 list entries represent 780 distinct production IDs. Three separate one-act titles map to production 490498, Relatively Speaking, counted once. Exclusions: 83 non-play/musical productions and 18 benefits. Final: 679 regular productions. Every retained entry has an opening date and performance count; no failed captures remain.
The production is the unit; a revival or return engagement with a different production ID counts separately. Genre comes from the season list. Benefit concerts are retained in the screening file but excluded because they are not regular production runs. No exclusions depend on length or review quality. IBDB's season definition is used without substituting Tony eligibility years or Broadway League reporting periods. Consequently season counts should not be assumed identical to those other frames.
We extracted opening/closing dates, performance and preview counts, and the count-as-of date from production headers. Closure dates are historical; document capture dates are not closure dates. Six runs remain open. Right-censored Kaplan–Meier medians are the first performance count where estimated survival is at most 0.5. This convention can differ from the arithmetic midpoint of two central observations. No confidence intervals are presented: this is a finite enumerated cohort, not a random sample. We report observed threshold attainment; every ongoing run already exceeds both thresholds used here.
Elapsed calendar days would count the pandemic closure, so performances are the main duration measure. Performances of multi-part works use IBDB's production-level convention; we do not infer admissions or audience-days. Limited engagements, changing theatre capacity, cast changes and operating costs are not controlled. Do not interpret the estimates as profits or a causal genre effect. Generalization outside these opening seasons requires additional evidence.
Files preserve the full season frame, exclusions, individual URLs and SHA-256 hashes of captured production pages. The collection and analysis scripts accompany the research package. Original publisher pages remain the sources for individual show facts; the new contribution is the joined longitudinal comparison.