jiu3 pronounced as jiu4
It is very prominently pronounced with a descending glissando. It is a distinct 4th tone, phonetically.
This is not true. Tone 4 is high and descending. It has no turning up, which distinguishes it from Tone 3. The turning in the plot is non-trivial.
As a single syllable, because jiu3 is often not said as a whole sentence, the tone might sound a bit weird. But it you say 好 with the exactly same pitch as jiu3 in the audio, then it feels natural.
linked speech feature or slip of the tongue? Can it still be a valid phonological 3rd tone? Or is it an accidental intonation mistake?
In the sentence, it's definitely a Tone 3 to a native ear. It's a valid phonological Tone 3; actually it's how Tone 3 usually sounds in connected speech, and even in non-connected speech in an informal style and for people not from Beijing or similar high-pitch regions, like the example 好 above.
Tone 3 is high/mid-high->low->mid-high. You can indeed see the little going up at the end of the pitch contour. It being short is not a problem. Neither is it not going up to mid-high. These are not its defining characters. What's important is the turning up, which you can clearly see from the plot.
On the other hand, sometimes the required tone contour is obtained together with the next syllable; this is usually the basis for 连续变调. The next syllable is Tone 1, by standard a high-level. Though in reality many people pronounce it not that high, especially non-Beijingers. (A high Tone 1 is a feature of Beijing Mandarin.) Together with 都, 久 completes a full high/mid-high->low->mid-high contour. (The automatically generate pitch line for 都 is not correct) This is why it sounds clearer in the sentence than in the single syllable that it's Tone 3.
RESPONSE TO COMMENTS @PapaSmurf
A break pitch line is most often from a Tone 3. Praat often fails to draw it correctly. Moreover, some people actually insert a glottal stop in the middle, making it more disconnected.
What makes it more complicated is that Praat sometimes gives a small connected rise at the end of a Tone 4. That is not part of the syllable, but more of a before-pause regression to mid-pitch with a more centralized version (schwa-shading) of the last vowel. More specifically, because Tone 4 ends low, the added part will go up. Tone 2 ends high, then the added part will down. This is a relaxed way of speaking; before pause and end of the sound the mouth returns to a neutral position and pitch to mid. It's especially obvious if you hear South-westerners speak Standard Mandarin; for other regions this phenomenon generally exists at a lesser degree. It disappears when people try to speak more carefully and with control, and it's also not likely to be heard from news broadcasters.
From the pitch line alone, it's sometimes hard to tell whether the turning is part of the syllable or not when the part after the turning point is short. Analyzing the vowel quality helps. Also, Tone 3 is low and twisted. The high part is not necessary but the low part is. For example, 212 could also occur as Tone 3 in connected speech, but never would something like 545 be considered as one. Tone 4 is high and sharp. It's more prominent with high/mid-high pitch and the change in pitch is larger than Tone 3 and also faster. Still, we usually confirms it by ears.
In the audio, 口语 only has emphasis of the 1st syllable. This reduced the length and clarity of 语. This includes a scale-down of the pitch contour. But the general shape retains.