The Tarantula Massive Binary Monitoring V. R 144: a wind-eclipsing binary with a total mass > 140 Msun

R 144 is the visually brightest WR star in the Large Magellanic Cloud (LMC). R 144 was reported to be a binary, making it potentially the most massive binary thus observed. We perform a comprehensive spectral, photometric, orbital, and polarimetric analysis of R 144. R 144 is an eccentric (e=0.51) 74.2-d binary comprising two relatively evolved (age~2 Myr), H-rich WR stars. The hotter primary (WN5/6h, T=50 kK) and the cooler secondary (WN6/7h,T=45kK) have nearly equal masses. The combination of low rotation and H-depletion observed in the system is well reproduced by contemporary evolution models that include boosted mass-loss at the upper-mass end. The systemic velocity of R 144 and its relative isolation suggest that it was ejected as a runaway from the neighbouring R 136 cluster. The optical light-curve shows a clear orbital modulation that can be well explained as a combination of two processes: excess emission stemming from wind-wind collisions and double wind eclipses. Our light-curve model implies an orbital inclination of i=60.4+-1.5deg, resulting in accurately constrained dynamical masses of 74+-4 and 69+-4 Msun. Assuming that both binary components are core H-burning, these masses are difficult to reconcile with the derived luminosities (logL1,2 = 6.44, 6.39 [Lsun]), which correspond to evolutionary masses of the order of 110 and 100Msun, respectively. Taken at face value, our results imply that both stars have high classical Eddington factors of Gamma_e = 0.78+-0.1. If the stars are on the main sequence, their derived radii (~25Rsun) suggest that they are only slightly inflated, even at this high Eddington factor. Alternatively, the stars could be core-He burning, strongly inflated from the regular size of classical Wolf-Rayet stars (~1Rsun), a scenario that could help resolve the observed mass discrepancy.