there is a blocksize limit, but it is subtle. Storing data in op_returns would actually decrease average blocksize since it is not discounted in segwit. This is the whole block weight concept introduced by pieter wuille. This is why inscription jpgs are included on the witness input script stack, to get the witness discount (an unintended spam side effect of segwit) The whole op_return narrative is weird, since if you want to store large images you would’t even use op_return, and the core v30 changes have nothing to do with that because data carrier settings don’t look at the witness stack. I think luke tried to PR a mechanism for detecting inscriptions and including that in the data carrier size, but it would have confiscated peoples money for transactions that looked like spam but were actually legitimate. this would have also created a cat and mouse game of various ways to match input script templates to try to get around every inscription variant, leading to an OFAC banning mechanism to bitcoin. nevent1qqsza8m029sye43ksllqtsxwey9a7cvt6cthxuxwy3yu4l2j8ytegacjc5t0p

You can get 2TB now for not more than 100 bucks and that would be sufficient for several years. So it's a non-issue.