Next-generation internet of things (IoT) applications need trillions of low-powered wireless mobile devices to connect with each other having ultra-reliability and low-latency. Non-orthogonal multiple access (NOMA) is a promising technology to address massive connectivity for 5G and beyond by accommodating several users within the same orthogonal resource block. Therefore, this article explores hybrid NOMA (HNOMA) for massive multiple access in the uplink scenarios due to its higher spectral efficiency. The HNOMA includes both power domain and code domain NOMA method due to diverse channel conditions in practice. We highlight that polar coded based data transmission can achieve higher reliability and lower latency in HNOMA-based wireless networks. Further, at the base station (BS), channel state information (CSI) of each link is not perfectly available or very complex to estimate due to non-orthogonal links. Therefore, we analyze and review the performance of uplink based system involving HNOMA transmission in the presence of imperfect CSI. Furthermore, we summarize some key technical challenges as well as their potential solutions in futuristic IoT applications using HNOMA transmission. Finally, we offer some design guidelines for HNOMA-based systems using deep learning approach to implement adaptive and efficient wireless networks.