Qt 6 now enables HTTP/2 by default, so Plasma’s requests to Google’s servers all attempt to use HTTP/2. It then fails due to Qt’s HTTP/2 client and Google’s HTTP/2 server disagreeing on the protocol exchange. AFAIK Qt’s HTTP/2 handler (qhttp2protocolhandler.cpp) throws a connectionError(PROTOCOL_ERROR, "HEADERS on invalid stream") as soon as Google sends the frame back, Qt’s QHttp2ProtocolHandler::handleHEADERS() rejects the response headers on the grounds that Qt has already decided the stream is closed for some reason. Sometimes it still gets an image either by caching it from something or a fallback to HTTP/1. Maybe it times out and then falls back too quickly leading to conflicts when the frame is sent back? IDK. All ik is when plasmashell gets an empty QPixmap you see a blank or fallback tooltip plus log spam from qt.network.http2.connection. At least that's what I think is happening from the cryptic ass strace. I found a recent bug report that included this same
protocol error elsewhere recently.
maybe I am wrong though. it's also not a super bad bug, the fallback works most of the time and nothing is crashing. Thats my guess as to why nobody is really talking about this very easily reproducible issue, but its annoying when u have a system set up to actually notify you when journalctl tells you of CRITICAL prio stuff.