On Fri, Feb 3, 2023 at 9:09 PM Andres Freund <andres@anarazel.de> wrote:
> Thinking about popen() suggests that we have a similar problem with COPY
> FROM PROGRAM as we have in pgarch (i.e. not as bad as the startup
> process issue, but still not great, due to
> procsignal_sigusr1_handler()).
A small mercy: while we promote some kinds of fatal-ish signals to
group level with kill(-PID, ...), we don't do that for SIGUSR1 for
latches or procsignals.