"Upon checking PERC H730P Mini, I saw that there was a note on the HCL recommending the HBA330 instead, any thoughts/considerations?"
H730P controllers were physically designed as a cache-backed RAID controller that were kind of slowly shoe-horned into providing a very different purpose which the HBA330 was designed for from the ground-up (passthrough-only, no cache, solid QD, less frills means less hassle) - while H730P with modern driver+firmware are *relatively* more reliable than they have been in the last few years I still wouldn't pick one over a HBA330. Plus they have added caveats e.g. no VMFS even just for logging on same controller as vSAN devices.
"As for the disks, what information is needed to confirm compatibility? (suspect model name is enough and then matching with vSAN guide PDF, see below:)"
Part-numbers can be more specific than Model and are also searchable via the vSAN VCG and/or if vendors list these but not full Model name:
VMware Compatibility Guide - vsan (click 'Build Your Own based on Certified Components.')
A colleague at GSS also made this which is very useful for quick searches by Model/Vendor/IDs:
"For all flash is an NVMe at the cache layer and SSD's at the capacity layer the recommended option?"
If the required performance and price-point are justified by the workloads requirements then yes go NVMe if possible - devices such as P4800X are surprisingly good value compared to anything even close to half the performance in an SSD for not massively different prices (other than maybe have to go a bit smaller). One thing to consider is balance with regard to Cache and capacity-tier performance capabilities e.g. putting NVMe in-front of multiple 4TB 7.2k SATA drives wouldn't really do the job much better than lower-end cache-tier certified SSDs of similar size.
"Any other hardware in terms of vSAN compatibility that needs to be checked?"
Nope, other than all other hardware is supported for the version of ESXi in use and your NICs and switches are well-capable of handling the inter-node load (especially if talking bigger All-Flash clusters with features such as RAID5/6, the beefier the network the smoother this will run).