The availability of non-volatile random-access memory (NVRAM) spurred interest in its potential as a foundation for high-performance, persistent data structures. Our work focuses on applying and evaluating the flat combining synchronization paradigm on persistent memory systems. We present a novel approach, named persistent flat combining (PFC), along with a simplified version called fast persistent flat combining (FPFC), which is more efficient but has more limited use cases. Our algorithms combine operations from multiple threads into batches, leveraging not only the well-known benefits of general software combining but also the unique characteristics of NVRAM to develop a high-performance, persistent resizable array together with a persistent memory allocator. Our experimental evaluation, conducted on Intel’s Optane persistent memory platform, demonstrates the effectiveness of our implementation and shows substantial performance improvements over traditional locking mechanisms and other state-of-the-art persistent software combining techniques. We show that at high concurrency levels, our approach outperforms detectable flat combining (DFC) by a factor greater than 10 and PBstack (a persistent stack based on PBcomb) by a factor greater than 2.