I don't know what exactly was previously wrong or what the devs changed. But, little history lesson:
In older clients, as we know, most haste items used the worn Effect: Haste method of granting haste. This wasn't the same as a haste buff or a worn haste effect with a specific % -- the amount of haste didn't come from the spell effect, but from a hidden field on the item. (This let them avoid having a different spell called "Haste" for each haste item, since they could just reuse the same spell with different hidden values).
If you think about it, you can see how having Effect: Haste work that way, while also existing side-by-side with rare, fixed-% worn haste effects was kinda odd. The client would have required separate code to handle both cases; having fixed-% worn haste effects at all seems redundant when the hidden field is there to use.
In later clients, the hidden field was done away with, and replaced with the displayed Haste: n% field (explicitly a different field -- a lot of the ancient item fields were relocated and/or changed in later overhauls, plus iirc the hidden haste field did double duty as the min proc/click level on non-haste items originally), and pre-existing haste items were transitioned to this, losing Effect: Haste entirely, if they used it. It's reasonable to think that the code to handle the hidden haste field might have been removed from the client around the same time, which would make the Effect: Haste stuff not work.
And I think you can see how that might leave fixed-% worn haste effects in limbo. Did they rip out the client code to handle that at the same time? But also: did worn effects like that have to deal with buff stacking rules? Did they ever even work correctly? etc. There weren't a lot of worn effect items like that in the classic game, and it was (as far as I'm aware) always vague what, if any, rules there were governing different spell effects when they were "worn".
On the other hand,
here's a quick video showing the backstab timer with and without the worn "Seething Fury" effect, in the Trilogy client (and straight from the old spdat.eff marked 2001-08-22 off the CD). About 7 seconds with Seething Fury, and the expected 10 seconds without. Not
exactly proof, but it's reasonable to think that if the Trilogy client handled worn fixed-% haste this way (or at least Seething Fury, could be hardcoded I guess), then whatever timeline-specific client version p99 would be on probably would have worked similarly.