I've never quite understood why 21 machines, 4 or 5 nodes and mixing versions and then feeding a mirror of reality into it from a purpose built block passing node can't be done. All I hear is "test net has to run a while and be reset" and " it only had one version at a time on it" um, not a valid staging platform?
Saw a lot of finger pointing at "witnesses should read the code" from folks who don't document the code, make the changes completely understandable or obvious and cannot even articulate or predict the effects of the very code they wrote (badly and undocumented along the way)
It sounds good to the uninformed to "blame the witnesses" for "not being diligent" but that excuse only seems to come from those with reasons to deflect fault off themselves.
You mean, like this guy, who was the STINC employee claiming to be in charge of the testnet?
https://steemit.com/steem/@inertia/re-ats-david-re-steemitblog-hardfork-20-what-to-expect-tomorrow-20180925t032945094z