Jumble is mainly designed for browsing content directly from a specific relay. Nowadays, there are special types of relays like community relays, curated relays, or language-based relays. When you browse these, it’s independent of who you follow — the data only comes from that single relay, which makes it much faster.
The following feed, on the other hand, fetches notes from the write relays set by all the users you follow. That means it needs to pull data from dozens or even hundreds of relays and then aggregate them, so it’s naturally much slower.