connect-standalone.sh runs one process that reads connector configurations from files on its command line and keeps source offsets in a local file. Nothing fails over and exactly-once sources (Kafka 3.3 129 +) are unavailable: fine for a laptop test or an edge agent. Kafka's FileStream example connectors left the default classpath in Kafka 3.2; copy libs/connect-file-4.3.1.jar into a plugin folder to use them.
connect-distributed.sh workers form a cluster through Kafka: configurations arrive over REST, offsets and configs live in topics, and tasks move between workers. Use it even for one worker in production; a lost container then loses nothing.