Nucleus replacement devices (NRDs) have potential to treat degenerated or herniated intervertebral discs (IVDs). However, IVD height loss is a post-treatment complication. IVD height recovery involves the nucleus pulposus (NP), but the mechanism of this in response to physiological loads is not fully elucidated. This study aimed to characterise the non-linear recovery behaviour of the IVD in intact, post-nuclectomy, and post-NRD treatment states, under physiological loading. 36 bovine IVDs (12 intact, 12 post-nuclectomy, 12 post-treatment) underwent creep-recovery protocols simulating Sitting, Walking or Running, followed by 12 h of recovery. A rheological model decoupled the fluid-independent (elastic, fast) and fluid-dependent (slow) recovery phases. In post-nuclectomy and post-treatment groups, nuclectomy efficiency (ratio of NP removed to remaining NP) was quantified following post-test sectioning. Relative to intact, post-nuclectomy recovery significantly decreased in Sitting (−0.3 ± 0.4 mm, p < 0.05) and Walking (−0.6 ± 0.3 mm, p < 0.001) coupled with significant decreases to the slow response (p < 0.05). Post-nuclectomy, the fast and slow responses negatively correlated with nuclectomy efficiency (p < 0.05). In all protocols, the post-treatment group performed significantly worse in recovery (−0.5 ± 0.3 mm, p < 0.01) and the slow response (p < 0.05). Results suggest the NP mainly facilitates slow-phase recovery, linearly dependent on the amount of NP present. Failure of this NRD to recover is attributed to poor fluid imbibition. Additionally, unconfined NRD performance cannot be extrapolated to the in vitro response. This knowledge informs NRD design criteria to provide high osmotic pressure, and encourages testing standards to incorporate long-term recovery protocols.