We investigate the joint optimization of automatic repeat request (ARQ) and MIMO diversity for multi-user wireless communication systems. Exploiting MIMO for diversity gain and ARQ for error correction, our goal is to better integrate MIMO and ARQ in the transceivers to improve multi-user uplink performance. Specifically, we design optimal linear precoders for ARQ retransmissions by multiple wireless uplink terminals equipped with MIMO transmitters and we rely on joint detection of multiple retransmissions at the receivers. To better utilize ARQ diversity, we derive a set of progressive linear ARQ precoders for flat fading multiple-access channels based on the criterion of maximizing the mutual information of multiuser (re)-transmissions. We propose two efficient iterative numerical algorithms to determine the optimal precoders.