Andrew

It got a bit ahead,try to reset again. Seems some frequency or delay is wrong that kernel reads block too early while flash controller is still erasing it.

Yours is far from destroyed. At two boots - does flash erase error appear at same address twice in a row indicating write protect or it nibblezs indicating clock issue

https://openwrt.org/docs/guide-user/troubleshooting/failsafe_and_factory_reset perform firstboot reset in failsafe mode. tell if message ``` jffs2: Newly-erased block contained word 0xdeadc0de at offset 0x0 0000000 ``` changes after. Y: speed N: ro Try v21.02.7...

If the errormessage changes after next reset it is a clocking problem.

You should not close issue which persists...

You should not doubt usability of roaming between bands. Most movile phones will keep sockets connected while roaming to extend range like considering aps with same name same 'network' Subjective...

There is no explicit nft flow package selected, but one shows in backt$ace. i suspect something else pulled it as a dependency and added some fancy default nftables rule. really...

Come back if you discover excess dependencies or conflicts.

Can you identify who pulled nf-conntrack module? It should be like same nf-conntrack or xt_conntrack via xt_bridge, wrong to have both required at once. You fount it in mt76 driver,...

You had modules from kmod-nft-offload in backtrace....