Sub-acute ruminal acidosis (SARA) is sometimes observed along with reduced milk fat synthesis. Inconsistent responses may be explained by dietary fat levels. Twelve ruminally cannulated cows were used in a Latin square design investigating the timing of metabolic and milk fat changes during Induction and Recovery from SARA by altering starch levels in low-fat diets. Treatments were (1) SARA Induction, (2) Recovery and (3) Control. Sub-acute ruminal acidosis was induced by feeding a diet containing 29.4% starch, 24.0% NDF and 2.8% fatty acids (FAs), whereas the Recovery and Control diets contained 19.9% starch, 31.0% NDF and 2.6% FA. Relative to Control, DM intake (DMI) and milk yield were higher in SARA from days 14 to 21 and from days 10 to 21, respectively (P < 0.05). Milk fat content was reduced from days 3 to 14 in SARA (P < 0.05) compared with Control, while greater protein and lactose contents were observed from days 14 to 21 and 3 to 21, respectively (P < 0.05). Milk fat yield was reduced by SARA on day 3 (P < 0.05), whereas both protein and lactose yields were higher on days 14 and 21 (P < 0.05). The ruminal acetate-to-propionate ratio was lower, and the concentrations of propionate and lactate were higher in the SARA treatment compared with Control on day 21 (P < 0.05). Plasma insulin increased during SARA, whereas plasma non-esterified fatty acids and milk β-hydroxybutyrate decreased (P < 0.05). Similarly to fat yield, the yield of milk preformed FA (>16C) was lower on day 3 (P < 0.05) and tended to be lower on day 7 in SARA cows (P < 0.10), whereas yield of de novo FA (<16C) was higher on day 21 (P < 0.01) in the SARA group relative to Control. The t10- to t11-18:1 ratio increased during the SARA Induction period (P < 0.05), but the concentration of t10-18:1 remained below 0.5% of milk fat, and t10,c12 conjugated linoleic acid remained below detection levels. Odd-chain FA increased, whereas branched-chain FA was reduced during SARA Induction from days 3 to 21 (P < 0.05). Sub-acute ruminal acidosis reduced milk fat synthesis transiently. Such reduction was not associated with ruminal biohydrogenation intermediates but rather with a transient reduction in supply of preformed FA. Subsequent rescue of milk fat synthesis may be associated with higher availability of substrates due to increased DMI during SARA.