How do I select a coverage factor — is k=2 always correct?

k=2 assumes the combined distribution is approximately normal AND the combined uncertainty is reliably estimated (large effective degrees of freedom).

Use the Welch–Satterthwaite formula (GUM G.4) to compute ν_eff. If ν_eff ≥ ~30, k=2 is fine. If a Type A component from few repeats dominates (say 3–5 readings), ν_eff can be small and k should come from the t-table: e.g. ν_eff = 8 gives k ≈ 2.31 for 95.45 %.

Also check shape: if one rectangular component dominates the budget, the combined distribution is closer to rectangular than normal and k=2 over-covers; JCGM 101 Monte Carlo gives the honest interval. ILAC-P14 requires the coverage probability (~95 %) to be stated, so document whatever k you use and why.