Hi @ARM_Kami,
The sw size limitations is not caused by the NOR Flash in this case, the boot process loads your sw into SRAM to run from there, so the limitation is given by the system RAM. This SRAM size will vary in the S32G model, this is mentioned as system RAM at the datasheet of the product.

[Page 5, S32G2 Data Sheet, Rev. 6, 12/2022]
The QSPI bus is mapped to can access up to an 512MB flash.
Also, something else to consider is there is some peripherals that need use of the system RAM to run
1. PFE
Some SRAM was reserved for PFE by default in the device tree of BSP, show as the following:

You can change the default SRAM memory region to DDR.
2. LLCE
If you used the CAN2ETH or ETH2Can application for LLCE, then they need to notice some SRAM allocation in the applications, usually they are arrays and could check their SRAM address with debugger such as Lauterbach.
3. IPCF
The share SRAM memory (include local remote address and their size) could be configured, default configure shown as the following:

4. HSE
The memory map HSE could be found in attached S32G2_Memory_Map.xlsx, shown as the following:

5. eMMC boot
You need to find "Application boot image header" from IVT header and check its field "RAM start pointer" and "Code length" and know the location and size of your A53 application that loaded in SRAM. Make sure no overlapping between A53 and M7 application.
There's one more note about it:

Additional comments, you could check whether your application had been loaded in SRAM completely firstly through debugger or other viewer tool for memory.
Also, at 30.13.1 QuadSPI boot [S32G2 Reference Manual, Rev. 8, February 2024] the following is mentioned:

[page 1255, S32G2 Reference Manual, Rev. 8, February 2024]
So, you need to ensure you application can be loaded in less than 500ms.
Let me know if you have more questions.