What is the best procedure to extract API from tablet matrices:
a) to add up to a defined volume (volumetric flask) or
b) to add a defined extraction volume to the sample (pipette),
if the tablet is based on mannitol and contains about 5% (m/m) insoluble excipients?

My personal preference is to add by pipet (for example: pipet 25mL into a 50mL vol. flask, add n accurately weightd tablets...)

In my lab itīs also common to add a defined volume. But do you know, why?

It's probably because the insoluble excipients would throw off the volume your API was dissolved in if you were to QS flasks with extraction solvent. It would be by a trivial amount in most cases, but if you have a horse sized pill with a low concentration API in a small vessel, it could pose a significant systematic abberation from the accurate result desired.

It's also a bit easier in terms of sample prep - drop tablet(s), add fixed volume, shake and/or sonicate, filter, shoot.

You are ultimately reporting all your data with respect to a defined volume of sample. The most accurate way to make up a sample with a controlled volume is to use a volumetric flask. It is simply fundamental good practice in a chemistry lab.

While it is not very likely to be important in your environment, assume for a second that you have two tablet making machines. One makes pills with a volume of 1 mL, the other one with a volume of 0.2 mL. Both are making tablets with a content of active ingredient of 20 mg.
If I use the approach advocated here and I dissolve the pills in a volumetric flask that will be filled to 50 mL, I get the same result for the analysis of both pills. If I use your approach - adding 50 mL of solvent to each pill, my analytical results will be different, and I have the QC department scratch their heads why the 1 mL pills don't pass specs.

I agree with you at least partly. But what if your pills are made of mainly insoluble excipients - as DR wrote, these excipients would throw off some volume - like placing stones in your flask....?