Bitmain was producing them minutes after a block change, and was producing-- prior to the blow up about covert AB-- upwards of 10% empty...
I guess I'll have to spend some time and check this ...
... since they've had Overt AB in the chips for years and no obvious large amount of Overt AB blocks on the network until they release the Overt firmware (before that was a 'small' amount by Dragon T1s which you can see the difference due to the different bits used)
I know James, with a little of -ck's help, did attempt to use (what is now called) AB in the miners after they came out ... but no idea if he used it long term, or if it was Covert or Overt
So if BM's been using Covert AB for years until the Overt firmware was released, the late empty blocks would be pretty obvious ...
I guess I can check that with a little code since I have the debug.log files for every server I've ever run since ... forever

... but I certainly don't remember seeing a lot of late empty blocks ...
... and the block header times mean nothing, it's when the blocks arrived, which I DO have that information from many server logs.