Flash droughts are rapid-onset hydroclimatic extremes driven by abrupt root-zone soil moisture depletion over sub-seasonal timescales. In monsoon-dominated regions such as India, their development is shaped by rainfall intermittency, seasonal soil-water recharge, atmospheric evaporative demand, and regional storage memory. However, the extent to which flash drought diagnostics depend on the choice of root-zone soil moisture (RZSM) product remains insufficiently quantified, limiting confidence in monitoring and early-warning applications. Here, we develop a percentile-based event-detection framework to diagnose flash droughts from two core features: rapid RZSM decline relative to local climatology and sustained soil-moisture deficit below a drought threshold. The framework is applied over India for 1981–2020 using four RZSM products representing hydrological modelling, land-surface reanalysis, and observation-constrained approaches: WaterGAP, H08, ERA5-Land, and GLEAM. Flash drought behaviour is quantified using event frequency, mean duration, onset speed, and cumulative severity. Product-dependent uncertainty is further evaluated using RZSM rank correlation, unbiased root mean square error, onset timing coherence, multi-product consensus probability, and structural spread.,The results show that flash drought frequency is the most product-sensitive metric. ERA5-Land and GLEAM generally detect fewer than 30 events across much of India, whereas WaterGAP produces localized maxima of approximately 60–80 events during the study period. In contrast, event dynamics are more consistent across products: mean duration is mostly 5–7 pentads, equivalent to about 25–35 days, and onset speed commonly ranges from 6 to 8 (percentile/pentad), with localized maxima approaching 8–10 (percentile/pentad). Mean severity is substantial across large parts of India, typically about 120–200 (percentile*pentads). RZSM rank correlations show stronger agreement over central and peninsular India, reaching about 0.7–0.9 for selected product pairs, but weaker agreement over northwestern India, the Himalayan region, and hydroclimatic transition zones. Onset-coherence diagnostics indicate closer alignment of WaterGAP and H08 with ERA5-Land, whereas GLEAM shows larger timing offsets and lower event coincidence. Regional temporal diagnostics further indicate that irrigated Indo-Gangetic regions, particularly Punjab–Haryana, exhibit distinct RZSM percentile evolution relative to rainfed regions, suggesting that human water management may modulate the apparent onset, depletion, and persistence of soil-moisture drought signals. Notably, consensus–spread diagnostics further reveal spatially selective agreement, with structural spread in frequency locally approaching about 42 events. These findings demonstrate that RZSM product structure strongly conditions flash drought diagnosis over India. While major flash drought-prone regions are broadly identifiable, event frequency, timing, and severity remain product dependent. By separating robust hydroclimatic signals from dataset-driven uncertainty, this study provides a physically interpretable benchmark for improving flash drought monitoring and early-warning confidence in strongly seasonal monsoon environments.