L&LD
Part of the Furniture
As a last resort, you may want to try the 9-step "nuclear reset" process by the well-known forum member @L&LD . Some people believe it, some people don't. But given the circumstances, I think you have nothing to lose.
Unless your router has a hardware problem, there's a good chance you'll recover it. If the "nuclear reset" doesn't work, then it's very likely your router is indeed defective, in which case you should return it.
Good luck!
Note that Step 2 should be using the link below for the method indicated for the router in use today.
Also, it is essential to note that doing the first few steps (resetting the router in as many ways as possible, and then flashing the (same) firmware you want to use/test with) should be repeated as many times as possible.
Then, only perform the M&M config as indicated in the link below.
M&M Config https://www.snbforums.com/threads/n...l-and-manual-configuration.27115/#post-205573
When you're in testing mode, do not further customize the router. Do not upload a saved backup config file. Do not plug in any USB drives, nor any LAN cables (you should be able to do at least preliminary testing with just the WAN connection and the WiFi), and particularly not any previously used USB drives that contained any amtm, or other, scripts.
If after a handful of complete resetting and re-flashing the same firmware cycles don't show improvements in the router's behavior, then this indicates it is more than likely a hardware issue at this point. At least for most individuals without any knowledge of further, low-level, probing/troubleshooting steps (such as myself).
I've 'saved' many routers using the method outlined above and in that link, and particularly, an RT-AC86U showing random issues (including not rebooting properly) was saved with an extended cycle of resetting (using all possible methods), then flashing the RMerlin firmware at the time. It did take over an hour just doing those steps endlessly, but that router is still in use today and the rebooting (and never starting up) issue was also solved since that process was performed on it too.