Just to clarify, you have generated a new contig level assembly using only PacBio reads, and you have some short reads available too that you haven't used yet? What's the contiguity and size of your assembly at the moment? Tens, hundreds, or thousands of contigs?

You can scaffold with other sources data too. Optical mapping, linkage maps, and since you've used long reads already, highly optimized ultra-long read sequencing could be used in place of Hi-C. However, they all come with different problems in generating appropriate samples. Here is a review with some methodologies and considerations.

The question about whether scaffolding is necessary really depends on what you want to use it for. For most types of analysis, I would say a highly contiguous assembly is helpful, but not a requirement. Though it certainly impacts the quality and robustness of your findings. QTL mapping for example, if you find QTLs related to a specific factor all clustering to a single chromosome is an interesting finding, but if you only have a contig level assembly you wouldn't have identified that result.