Audio Ducking - Dip - Podcast - Apple Watch Ultra

I was listening to a podcast through my apple watch ultra and every time there's a longer pause (silence, nothing else in the background) the volume goes down and the next word gets cut off for a syllable.

Probably the threshold is 0.4 - 0.5 sec. It is Annoying and distracting. I have to guess the first word, after a break, and reconstruct the sentence.

I've verified this by listening to the exact same podcast, exact same sequence on a PC, it's all level, no dips.

I hope it's not set on purpose, but it's not the first time Apple would do something because they know better, big brother style. So Apple, get rid of it. It's stupid. Destroys the experience.