Just logged in to bump this up as helpful. Same issue happened with us. Increasing the max size turned this from an outage that would recover in 30 or so minutes, to an outage that took hours to recover and took down all our nodes.

Our fix was to lower truncation_limit in the Integration Instance record from 500 all the way down to 100, since that determines how many records get pulled in per "page." (Though, max_delta_days parameter like Bla_ena mentioned may help as well).

We've since crashed again after doing a plugin update, but fortunately that only happened in test and we were able to disable the integration. So... be careful updating to 12.8.4 and make sure to do a test run in a lower environment.