Well, I fear this won't help you (cause 1. at time I only enabled notice loglevel , 2. router won't crash on reboot if I had not left ssh enabled before rebooting), but here it is (see attached)
It's most propably the same issue I first noticed on .19a1:
You only have to do a nvram reset, reboot device, then configure as ap mode (setup wizard) with IP assigned by another DHCP server, reboot again, start with system administration page and e.g. change the following:
- Format JFFS partition at next boot: yes
- Enable HDD Hibernation: yes (5min)
- DST time zone change: (doesn't matter)
- Disable LEDs: yes
- Enable SSH: LAN only
- Enable SSH Brute Force Protection: yes (don't know anymore if I had this one enabled or not; perhaps this is the cause of my trouble, malfunctioning?)
Afterwards I had set some nvram settings (disabled NAT filters, upnp, WPS, STP, WRS, set loglevel and client customlist) via SSH, but from my POV these settings I changed doesn't matter at all. Then force a reboot, leaving ssh enabled and for sure - you won't get access to the device anymore (even not responding to icmp echo requests anymore; totally unaccessible by IP anymore even if you turn power off and back on) - all in all if I had left SSH enabled before rebooting. Otherwise: If I had disabled SSH after pushing nvram settings and then forcing a reboot everything is good: successful reboot, normal operation.
Maybe I've to submit you my config in order to be able to reproduce this issue?
HTH