Decades of observation in natural plant populations have revealed pervasive phenotypic selection for early flowering onset. This consistent pattern seems at odds with life-history theory, which predicts stabilizing selection on age and size at reproduction. Why is selection for later flowering rare? Moreover, extensive evidence demonstrates that flowering time can and does evolve. What maintains ongoing directional selection for early flowering? Several non-mutually exclusive processes can help to reconcile the apparent paradox of selection for early flowering. We outline four: selection through other fitness components may counter observed fecundity selection for early flowering; asymmetry in the flowering-time-fitness function may make selection for later flowering hard to detect; flowering time and fitness may be condition-dependent; and selection on flowering duration is largely unaccounted for. In this Viewpoint, we develop these four mechanisms, and highlight areas where further study will improve our understanding of flowering-time evolution.