The danger if that happens - the older (i.e. current generation) HomePods passing their requests up to Apple cloud servers for processing - is that either immediately or over time Apple might gradually allocate less and less cloud compute resource to processing requests originating from older HomePods in order to (both of these in Apple's interest not the consumer's) help keep Apple's server costs within a defined budget and incentivise users of older HomePods to upgrade because the consequence will be those users gradually seeing the response times to their queries increase. I fully accept that I could have been imagining this but I really do think that I have seen this effect in action with Apple already.
I got my first Apple Watch in the pandemic and use Siri for simple stuff like timers, device control etc. My first Watch had no real on-device AI processing capabilities and then during one launch (the Series 9 with the new S9 chip I think) one of the things Apple made a big deal of at the launch event was that the new S-Series chip could now process some requests entirely locally thus reducing response times. I upgraded and my immediate impression was that the response times between my old and new watch were pretty much identical. I did however keep my old watch as well for about 18 months and used it occasionally and over time I did notice the old watch getting noticeably slower than the new one in processing Siri requests.
Was that a real effect or my imagination? If real was that evidence of Apple gradually deprioritising the processing of Siri requests from old watches as a nudge to users of pre-S9 watches to upgrade or was it just different levels of software optimisation? I have no idea so you can essentially file this as a conspiracy theory but it does seem at least plausible to me, both for older Watches and potentially for existing HomePods in the year or two after the new ones are released.