On news platforms, users may be unable to access updates for an extended period or may wish to review a certain topic that has been progressing for quite some time. Experts writing articles on current developments, along with creating multiple timelines is a laborious task. These use cases highlight a need for automated timeline summarization (TLS). This motivated us to explore news timeline summarization that utilizes Natural Language Processing techniques, Machine Learning, and Large Language Models (LLMs). In this paper, our contributions are a review of recent research advancements in TLS that showcase the capabilities of LLMs, such as zero-shot, few-shot, and retrieval-augmented generation for generating coherent and reliable summaries. We also highlight key challenges in TLS, including the difficulty of maintaining joint representation of all topic-specific articles, weak or uncertain causal links between events, and redundancy due to adjacent dates within timeline referencing the same events. Additionally, we outline key research directions such as multi-timeline summarization based on the interest and perspective of individual user or a user segment, and the collection of high-quality articles that constitute topical relevance, broader domain coverage, and have strong inter-event links.