Telepresence robots enable remote and embodied teamwork. However, most studies focus on dyads. We compare a remote operator collaborating with either one or three co-located partners (1–1 vs. 1–3) in a treasure-hunt-style mock gallery with tasks that vary in coupling (forced vs. visually distributed). Across eight 1–1 and eight 1–3 sessions, we coded episodes for operator bids, uptake, repair triggers (audio/visibility/reference), and repair outcomes, and triangulated with post-activity interviews. We report three findings: (1) when work is locally doable, the operator becomes socially “optional,” especially in 1–3; (2) teams use similar repair strategies, but repairs reach closure less often in 1–3; and (3) unrepaired mishearing and visibility problems quickly cascade into subsequent ignores. We outline implications for re-coupling distributed work, reducing visual/audio grounding costs, and supporting operator entry into multi-party talk.
