The feeding of high-grain diets to dairy cows commonly results in lowered pH and ruminal dysbiosis, characterized by changes in absorption dynamics of short-chain fatty acids (SCFA) across the reticuloruminal wall, epithelial function, and the epithelial bacteria community structure. Therefore, the present study evaluated the effect of high-grain feeding persistence on the absorption kinetics of reticuloruminal SCFA, gene expression in the rumen epithelium, and the associated shifts in the epithelial bacteria in cows recovering from either a long-term continuous high-grain feeding model or a long-term transient high-grain feeding model. In a crossover study design, 8 nonlactating Holstein cows were fed 60% concentrate either continuously for 4 wk (continuous) or with a 1-wk break in the second week of the high-grain feeding (transient). After the high-grain feeding, all animals were fed a diet of 100% forage (recovery) for an additional 8 wk. Rumen papilla biopsies and SCFA absorption measurements were taken at the start of the trial (baseline), after the 4-wk high-grain feeding (49 d), after 2-wk recovery forage feeding (63 d), and after 8-wk recovery forage (105 d). Absorption of SCFA was determined in vivo using the washed and isolated reticulorumen technique. Rumen papillae biopsies were used for adherent bacterial DNA and host RNA extraction. The epithelial bacteria were determined using Illumina MiSeq (Microsynth AG, Balgach, Switzerland) sequencing of the 16S rRNA gene. No significant effects of the high-grain feeding model were seen for bacterial diversity. However, bacterial diversity increased with time spent in the recovery forage feeding period regardless of feeding model. The relative abundance of Acidobacteria phyla and Acetivibrio spp. increased when animals were fed a transient high-grain feeding model. A trend toward increased CLDN4 expression was observed in the continuous model. Furthermore, there were interactions between feeding model and sampling day for gene targets CD14, DRA, NHE2, NHE3, and MCT2. When comparing length of recovery, in the continuous model increased relative absorption of SCFA was sustained at 63 d but dropped to baseline measurements at 105 d. A similar pattern was found with the transient model but it did not reach significance. The only gene target that was found to significantly correlate to relative absorption of SCFA was DRA (correlation coefficient ≤ −0.41). Whereas, genera Alkalibaculum, Anaerorhabdus, Coprococcus, and Dethiobacter all showed positive correlations to gene targets for pH regulation (NHE2 and NHE3) and SCFA uptake (MCT1) but negative correlations to SCFA absorption. We conclude that while the rumen absorption and epithelial bacteria were able to recover to baseline levels after 8 wk of forage feeding, the time needed for re-establishment of homeostasis in host gene expression is longer, especially when high-grain feeding is interrupted.