That’s generally the case — and depending on the hosting system, that could change too. The really huge companies in the space (Art19, Stitcher/Midroll, Wondery, Megaphone, Podbean) are all doing dynamic ad insertion on different scales — but given the way the RSS feeds are constructed, I don’t know specifically how the dynamic aspect works for each provider.
I do know that early on, companies who did this stuff (like NPR, who was really early to this game and has been doing some type of dynamic ad-insertion for more than a decade) often had separate versions of the audio file. So you could have an ad-free version of the file (maybe for users in certain geographic regions, if an ad-buy was specifically targeted) and then other versions of the file with ads built-in — and an algorithm would determine what version a user accessed based on a variety of parameters. If that’s the case, yeah, you would probably be able to figure out if multiple versions existed and then request the ad-free version from CDN or whatever.
At this point, it’s my understanding that most of the dynamic insertion is happening server-side — where ads are inserted based on campaign timing and maybe geo-targeting) at the point of the download or stream request (though I imagine even here, most of what is served is coming from a CDN, I just mean the initial request). That way, the download/stream always has “current” ads, even if the show itself is years old. (I don’t know how often they update those different versions — I would imagine you have 6-week campaign as an example where all episodes get an insert and are then cached on the CDN and then after the campaign ends a new file is generated and cached). That isn’t to say you couldn’t potentially still get access to a copy of a file without inserts — but I bet it would be a lot more difficult.