The hfs+ implementation itself is slow in Linux, either in kernel or via fuse. But if speed isn't an issue have at it. You will still continue to have the read only on reboot issue - it's a safety setting in the hfs implementation to stop mac drives being corrupted.
It's your device so do as you will - but using hfs+ in linux for anything beyond interoperability on a dual boot mac is a terrible idea. And yes, I have done exactly what you are trying to do (though on an x86 server) and had this same issue with random RO on the drive. You're likely to get file corruption as well.
Is there a way to disable this? This is a test device. And, I would like to see what the breaking point is.
Can you elaborate a bit more on how you built and implemented the ext4 drive? Just incase I wanted to correctly prevent this.
Currently MacOS (Big SUR -> Ventura) creates an APFS bundle natively. Which made me preconfigure the drive to use HFS+ then build the sparsebundle over SMB.
macOS supports timemachine over smb natively, you don’t need to create a sparse bundle or do any prep, beyond configuring samba with the Time Machine supporting features.