The log entries looks to be related to PCIe Base Address Register (BAR). So it unlikely to be bad RAM hardware if it is PCI BAR.
I don't think that it is a memory leak. My guess the "Unable to free" gives that impression. It is unable to free memory locations that it failed to add.
2020-10-12T05:38:48.017Z cpu0:524288)WARNING: PCI: 320: 0000:00:12.0: Failed to add BAR[0] (MEM64 f=0x4 0x4000006000-0x4000007000) status: Limit exceeded, parent: \_SB_.PC00
2020-10-12T05:38:48.017Z cpu0:524288)WARNING: PCI: 184: Failed to remove Non-PF mem (0x4000006000 - 0x4000006fff): Limit exceeded
2020-10-12T05:38:48.017Z cpu0:524288)WARNING: PCI: 453: 0000:00:12.0: Unable to free BAR[0] (MEM64 f=0x4 0x4000006000-0x4000007000): Limit exceeded
Does this host causing grief have VM(s) that have PCIe passthrough configured? Perhaps someone else is trying a PCIe device passthrough in a VM and knowingly/unknowingly caused a host crash while trying to boot up the VM and this person just keeps quiet.
If you have VM(s) that have PCIe passthrough, make sure the host is using UEFI and have MMIO above 4GB. The VM(s) also need to be using virtual EFI and have the 64-bit MMIO enabled.
Look at this https://kb.vmware.com/s/article/2142307 for reference.
