Training in parallel on more then one GPU is pretty hard. Distributing the training over many random computers and adding at least some kind of validation, seems to kill the performance. So it would always be easier to rend cloud GPUs.
Not true. Most RxDB users use the dexie.js storage which is for free: https://rxdb.info/rx-storage-dexie.html
Also you should compare it to the alternatives like firestore or AWS datastore which are not even open source partially and cannot even be used with a self-hosted server.