I tried the process in Wavelab.

Here is my file:

Isn't that what you'd end up with if you digitized a 30 minute cassette 
if the recording speed had gone slower during the recording?
I didn't use an analog recorder for this file, I made a demo file in the 
At the end of the recording the pitch is 16 semitones higher than in the 
The test file duration is 29'26".

Here is the result of the WaveLab "Pitch Bend" process:

The result is 45'26", which is quite close to my original starting 
audio, 45'06".
Could have tweaked it.

The Wavelab process doesn't seem to be linear. The pitch lowered in the 
of the sequence, but was ok in the end. I adjusted the contour to 
correct this.

Untick the "Preserve duration" box in the Pitch Bend window.
The maximum range of the WaveLab Pitch Bend is 36 semitones.

If you just need to make the interwiew easier to listen to, this seems 
to work.

In practise, you had better first find by ear and using the speed adjustment
the needed speed correction at the end of the recording and in a couple 
of spots
along the way. The recording speed most likely hasn't dropped linearly 
and it
drops steeper towards the end. The cassette recorder speed control has 
tried to
keep it up, but when the battery voltage drops, the speed will also slow 

Did I miss something?

An automated process would surely speed up the job.