I’ve experienced this twice, and both times, the system was in OpenWRT failsafe mode and had an SSH server running, so I could log into it. From there I was able to determine that one of the mtd devices had been created but wasn’t completely formatted, causing the upgrade scripts to continuously loop, thinking the partition is ready for data when it isn’t, and subsequently timing out after a couple of minutes of waiting for it to come ready and rebooting, over and over.
If the lights are cycling briefly every few minutes, and when they stabilize to just the power light you are able to SSH into the box (but it later kicks you off), you may be experiencing exactly this, and it’s easy to fix. Let me know if you can get an SSH connection at any point (you may have to try several times over several minutes), and if you can, I’ll dig out my notes and send them to you. The fix is as easy as clearing the partition with one command, and then the thing fixes itself from there. There are actually a couple of versions of this problem and fix buried in forum threads here.