Second of all, revisions are incredibly common with BLS numbers, it's normal. They are trying to balance accuracy with speed, and this is how they do that.
They survey about 50,000 companies out of the 12 million registered US companies. Since the pandemic, response rates on their surveys have gotten worse. It is now about 50% at the end of the first month, slowly rising to about 90% two months later, so they update their results as the surveys get completed. The waiting is especially important because the speed of survey returns turns out to not be randomly distributed, the bigger the business, the faster it returns a survey, so small businesses are more likely to respond later. The BLS attempts to counter for that in their initial reports by inflating the small business returns that do come in early enough to account for their under-reporting, but the later returns are just incredibly valuable for getting accurate data.
Even with all of that, under normal economic circumstances revisions generally are small (they exist, but generally they cancel out thanks to the BLS inflation of the small business numbers, some of the businesses are higher than expected, some are lower, it all cancels out). Large revisions, either up or down, are generally associated with the economy changing course (lots of small businesses miss in the same direction, pretty good sign that the economy is changing)- large revisions down are often seen as a leading indicator of a recession, and large revisions up are often seen as a leading indicator of an economic improvement.
This is all true at least in the previous regime, where the BLS was trusted to be non-partisan. That reputation has definitely taken a hit; we will have to see how much trust remains in that organization going forward.
She was fired specifically for "cooking the numbers." An attempt to bring BLS methodology into question is certainly more productive than what Trump did, but it has little to do with Trump's state of mind when he fired her.
