Transit surveys have revealed a significant population of compact multi-planet systems, containing several sub-Neptune-mass planets on close-in, tightly-packed orbits. These systems are thought to have formed through a final phase of giant impacts, which would tend to leave systems close to the edge of stability. Here, we assess this hypothesis, comparing observed eccentricities as measured in systems exhibiting transit-timing variations (TTVs), with the maximum eccentricities compatible with long-term stability. We use the machine-learning classifier SPOCK (Tamayo et al. 2020) to rapidly classify the stability of numerous initial configurations and hence determine these stability limits. While previous studies have argued that multi-planet systems are often maximally packed, in the sense that they could not host any additional planets, we find that the existing planets in these systems have measured eccentricities below the limits allowed by stability by a factor of 2–10. We compare these results against predictions from the giant impact theory of planet formation, derived from both N-body integrations and theoretical expectations that in the absence of dissipation, the orbits of such planets should be distributed uniformly throughout the phase space volume allowed by stability. We find that the observed systems have systematically lower eccentricities than this scenario predicts, with a median eccentricity about 4 times lower than predicted. These findings suggest that if such systems formed through giant impacts, then some orbital damping must have occurred during or after formation, perhaps due to interactions with the natal gas disk or a leftover population of planetesimals.