We propose to conduct a prospective cohort (observational) study of 150 participants with moderate to severe obstructive sleep apnea (OSA) unable to tolerate positive airway pressure who are undergoing drug-induced sleep endoscopy (DISE), including measurement of upper airway closing pressure (Pclose), and tissue-repositioning soft palate surgery. Before and 6 months after surgery, we will measure OSA severity (apnea-hypopnea index) with sleep studies (polysomnograms). Using our recently-validated polysomnography-based signal processing algorithm, we will systematically assess the underlying mechanisms of OSA (traits) and airflow shape (consistent with palate obstruction). The expected 90 participants without resolution of OSA after surgery (surgery failures) will participate in an experimental randomized crossover study of acetazolamide (1 month) and acetazolamide/eszopiclone combination (1 month). Polysomnograms will be performed with each treatment, with algorithm-based determination of traits. The research goals are: (1) to define how mechanistic traits modify effectiveness of anatomic treatment (surgery) and (2) to examine the therapeutic efficacy of surgery in combination with endotype-directed (especially loop gain- and arousal threshold-directed) treatment. Aim 1: To determine the mechanistic factors modifying the impact of soft palate surgery on airway collapsibility. Hypothesis: tissue-repositioning palate surgery will decrease airway collapsibility, with the decrease greater with palate-only obstruction (based on DISE and non-invasive airflow shape analysis). Aim 2: To evaluate the mechanistic factors modifying efficacy of soft palate surgery on OSA severity. Hypotheses: surgery efficacy will be independently associated with palate-only obstruction (DISE) and the following prior to surgery: lesser airway collapsibility, lesser loop gain, and greater arousal threshold. Aim 3: To evaluate the efficacy of combining surgery with loop gain and arousal threshold interventions. Hypotheses: In surgery failures (non-responders), lowering loop gain (acetazolamide) will lower OSA severity (expected response rate = 40%); efficacy will be associated with lesser postoperative (prior to medication) airway collapsibility and greater postoperative loop gain. Lowering loop gain and raising arousal threshold together (acetazolamide/eszopiclone) will further reduce OSA severity (expected response rate = 60%); efficacy will be associated with lesser postoperative airway collapsibility and the combination of greater postoperative loop gain and lesser postoperative arousal threshold.