I’m offering another update.
Vacuuming and setting up a vacuum routine made a difference - peaks in resource usages are as “tall” as they were before, but at least they are not as “wide” as they have been last month.
This isn’t solving the situation.
Operations involving large sections of the table structure (accessing the user profile, or moving posts, for instance) are sometimes met with a 502 error.
Wanting to preserve the current forum structure, we are going to move to a dedicated machine as soon as possibile.
Thanks for all the help so far.