The network process used to be robust, but nowadays we have multiple regressions resulting in semi-frequent network process crashes, most notably bug #260962 and https://gitlab.gnome.org/GNOME/libsoup/-/issues/313. When the network process crashes, networking breaks in all existing tabs: * It's no longer possible to post comments or upload anything GitLab. * Often, but not always, I'll discover that I've been involuntarily signed out of GitLab, as if my cookies had been deleted. * Slack silently stops displaying any new messages with no indication that anything is wrong unless I try to send a message, which will stall and never post * Element sometimes complains that its database has been unexpectedly closed and forces me to reload the page, but other times it just silently stops displaying new messages without any indication of a problem. Unread message indicators stop updating. It will also silently discard new messages that I post while displaying them as if they were sent. We should to plan for network process crashes to happen and ensure we recover gracefully.