"Acceptable" is arguable and subjective. Every day when I open multiple web applications on my phone, I tacitly accept the performance, as does everyone else who does the same. But, of course, we would also all love everything to be faster and for our batteries to last longer. Those desires are being attacked from different angles with different trade-offs all the time. I'm not at all convinced that the pace of improvement has leveled off, but that's just my gut feeling, and I would be interested in seeing analysis. I'm also not convinced that, assuming evolutionary improvement has indeed leveled off, there isn't still plenty of potential for new techniques that bear more fruit. One area of research I'm aware of targeting further improvement to performance is increasing concurrency, which the servo project[0] is exploring. Perhaps things like asm.js and webassembly will also make an impact.
I'm not a complete web apologist and I'll freely admit that you may well be right that it is inevitably a resource hog, but it is an incredibly useful platform, and I think giving up on it would be "throwing the baby out with the bathwater".
(I also find it pretty crazy that things like Slack and Atom spin up a full web rendering environment, and I would rather see us have better tools for making nice cross-platform desktop applications where they make sense.)
[0]: https://github.com/servo/servo