Reliability of a PG 6541B gas turbine (Unit No. 03) at the Shazand Petrochemical Complex was assessed using time-between-failures (TBF) data collected after the 2019 major overhaul. Thirty uncensored TBF intervals (31 forced-outage events) were recorded over 1692 operating days (June 2020–January 2025). Treating the turbine as a repairable system and adopting a system-level functional renewal (as-good-as-new) approximation, the empirical mean time between failures (MTBF) was 56.40 days (95% confidence interval (CI): 40.63–83.59 days). Comparative distribution fitting (maximum likelihood estimation) selected the two-parameter Weibull model as the best parametric summary of inter-failure times, with shape β = 1.74 (95% CI: 1.32–2.21) and scale η = 1084 days; in this repairable-system context, η is interpreted as the scale of the inter-failure distribution and not as a deterministic lifetime. To address conceptual ambiguity, MTBF is defined as the arithmetic mean of observed TBFs, whereas the Weibull-derived mean inter-failure time (denoted as MTTF in this paper) represents a model-based expectation under the stated renewal-type approximation and can differ substantially in right-skewed small samples. Goodness-of-fit tests (Anderson–Darling and Kolmogorov–Smirnov) supported the adequacy of the selected model. Trend diagnostics showed no statistically significant monotonic trend in the failure process (Kendall τ = −0.097, p = 0.46; Laplace U = 0.761). Sensitivity checks indicated robustness to extreme observations (β changed from 1.74 to 1.68 when the shortest and longest TBFs were excluded). This work contributes an uncertainty-aware industrial case study based on screened long-term field records, providing defensible reliability indicators and interpretation guidance for maintenance planning in petrochemical gas turbines.