GJ 436b might be the prototype of warm Neptunes that have undergone late migration induced by an outer companion. Precise determination of the orbital architecture of such systems is critical to constraining their dynamical history and evaluating the role of delayed migration in the exoplanet population. To this purpose we analyzed the Rossiter–McLaughlin (RM) signal of GJ 436 b in two transits – recently observed with ESPRESSO – using three different techniques. The high level of precision achieved in radial velocity (RV) measurements allows us to detect the deviation from the Keplerian orbit, despite the slow rotation of the M dwarf host (v sin i_* = 272.0_−34.0^+40.0 m s⁻¹), and to measure the sky-projected obliquity (λ = 102.5_−18.5^+17.2°). The Reloaded RM technique, which allows the stellar RV field along the transit chord to be analyzed, yields λ = 107.5_−19.3^+26.6° and v sin i_* = 292.9_−49.9^+41.9 m s⁻¹. The RM Revolutions technique, which allows us to fit the spectral profiles from all planet-occulted regions together, yields λ = 114.1_−17.8^+22.8° and v sin i_* = 300.5_−57.0^+45.9 m s⁻¹. The consistent results between these three techniques, and with published results from HARPS/HARPS-N data, confirm the polar orbit of GJ 436b and support the hypothesis that its origin lies in Kozai migration. Results from a joint RM Revolutions analysis of the ESPRESSO, HARPS, and HARPS-N datasets (λ = 113.5_−17.3^+23.3°; v sin i_* = 293.5_−52.2^+43.7 m s⁻¹) combined with a revised stellar inclination (i_* = 35.7_−7.6^+5.9° or 144.2_−5.9^+7.6°) lead us to constrain the 3D obliquity Ψ to 103.2_−11.5^+12.8°.