The hidden slippage of routing multi-hop swaps during chain congestion

We talk a lot about front-end speed, but the real silent killer for larger orders on non-custodial networks is how aggregators handle path routing during sudden network congestion. When block space fills up, multi-hop routes often get front-run or suffer from stale price impact data before confirmation. Are you guys splitting orders manually across different routing paths or relying strictly on single-path aggregators when moving size? Looking to see what execution setups actually hold up.