Correct. I still use Lazy Newb Pack with DFHack, DT, Announcement Filter, and the Phoebus tile set, but gave up Workflow only. If DFHack is set up as I usually dol, I should still use performance tweaks and pure bug fixes.
A theory I had was that the amount of available jobs somehow pushed open a leak so jobs still show up in the jobs list, but are lost from the invisible presumed available jobs list (a behavior that would be quite possible with leaky multi threading, but DF is single threaded, AFAIK). I've reduced the set of invisible jobs (i.e. jobs that are available, but never listed until taken) by aggressively using QS with limited size feeder stockpiles (every empty stockpile tile generates a "haul here" job, if I understand things correctly). However, if invisible job reduction has any effect or not is completely unknown. I haven't had any "undug tile" problems even when designating fairly large areas for digging (but the recollection is not to be trusted, because once you hit the massive job loss problem, you tend to "correct" any suspected cases without thinking about it and without waiting to confirm it's a lost job, not just a tardy allocation).
I tend to run my dorfs ragged, getting them to do something at all times (harder with the new job scheduling, because the are now reported as "idle" between each job), with more than 0 unemployed dorfs a management failure. Seems to have a social cost, though, and they're extremely bad at forming relations and, further on, generating dorflings.