Explain stale-while-revalidate caching on the client. How do you choose cache times and when do you invalidate?

Stale-while-revalidate means: show the cached data instantly, refetch in the background, and swap in fresh data when it arrives. The user never sees a spinner for data they have seen before.

TanStack Query has two settings for this:

• staleTime is how long data counts as fresh. During this time there is no refetch at all.
• gcTime is how long unused data stays in memory before it is removed.

I choose staleTime per resource. It depends on how fast the data changes and how much stale data hurts. Examples:

• A country list can stay fresh for hours.
• A user profile can stay fresh for a minute.
• Stock levels or a chat list need close to zero, with push updates on top.

Refetch triggers, like window focus and reconnect, cover users who come back to a tab.

I invalidate on mutations. For example, after creating a comment, invalidate that post's comment list, not everything. Hierarchical keys such as ["posts", id, "comments"] make targeted invalidation easy, because you can invalidate one branch of the tree.

For data that other users change, I combine short stale times with push events that invalidate specific keys.