Some of the pages load quite fast. The particular one you linked does not. I'm happy to give them the benefit of the doubt that there's a few rough edges that could be improved, and that the performance isn't a systemic problem.

The topic is the software, not how the hosting deployment performs under HNHoD.

it took 3.89 seconds for me just now even though the hnhod has presumably evaporated hours ago

this post-a-comment page took 1.27 seconds, presumably in large part because i'm in argentina. but that log page took about 2.8 seconds to load just the html, so there's still a painful performance problem there that isn't related to argentina

I think the idea is rather that you would set it up on your own host. What you're mentioning could likely be alleviated by a properly configured caching reverse proxy.

maybe

a lot of backend software doesn't provide cache invalidation signals

a lot of caching reverse proxies don't support cache invalidation, or don't support it efficiently

you can't always prerender all possible views

and this is the kind of situation where a slightly out-of-date page could be a huge hassle or even cause a major outage; imagine pushing a branch to production because you thought it didn't have your latest buggy commit in it

it probably isn't going to be any faster on my own host

The first two words on the page are "Hyper Performant". Pointing that out as baseless hype seems fine.

It's running on a tiny dedicated server in the Netherlands with an Intel Atom C2750 (2013) so that's something to keep in mind. By default the log page will return 100 commits which is more than most forges return. Thanks for pointing out the slowness!

There is a good performance comparison of forge performance here: https://forgeperf.org. Ayllu isn't feature complete yet so I haven't run those benchmarks against it.