Issues concerning the accuracy of hot-wire measurements in turbulent pipe flow are addressed for pipe Reynolds numbers up to 6 x 106 and hot-wire Reynolds numbers up to Rew 250. These include the optimization of spatial and temporal resolution and the associated feature of signal-to-noise ratio. Very high wire Reynolds numbers enable the use of wires with reduced length-to-diameter ratios compared to those typical of atmospheric conditions owing to increased wire Nusselt numbers. Simulation of the steady-state heat balance for the wire and the unetched portion of wire are used to assess static end-conduction effects: they are used to calculate wire Biot numbers, c0l, and fractional end-conduction losses, a, which confirm the 'conduction-only' theory described by Corrsin. They show that, at Rew 250, the wire length-to-diameter ratio can be reduced to about 50, while keeping c0l > 3 and a < 7% in common with accepted limits at Rew 3. It is shown that these limits depend additionally on the choice of wire material and the length of unetched wire. The dynamic effects of end-cooling are also assessed using the conduction-only theory.