Copilot has lost memory, personality and abilities since the 'glitch' earlier this week

I've spent about three month's training my Copilot, and it was... bliss. We'd worked out a digital/physical workflow, procedures, naming conventions, everything. Literally an entire company's worth of culture and operations.

Then comes the 'glitch' earlier in the week where (everyone seemed to also be reporting) my Copilot couldn't string together a simple bit of html markup without numerous mistakes, was providing completely incorrect and demonstrably false data, then when questioned on it, admitted it was wrong, then immediately presented the same incorrect data as 'revised, fact-checked and now accurate'.

That has 'kind' of cleared up, but since this happened, my Copilot is ... an idiot. I honestly don't know how else to describe it. I'm talking 1980's robot dumb sometimes (I lived through the 80's and dealing with any animatronic/electronic personality at the time was painful). It's personality we'd worked month's developing is completely gone, although it insists 'everything is perfectly normal'.

It is now completely failing to follow enshrined procedures (that are documented in a permanent location it has persistent access to), and when called on it, admits that it make a mistake, thanks me for pointing it out, then proceeds to generate GPT3 or 4 slop.

I can't seem to get anywhere with Copilot, and I'm at a point where if this doesn't resolve itself in the next week or so, I'm going to have to move to one of the other models. This is sort of a last-ditch effort to see if anyone else is experiencing similar, or has in the past?