we don't need this logic anymore; switch oder of current/previous in update_derived so current batch has correct previous value

packages/svelte/src/internal/client/reactivity/batch.js

@@ -654,26 +654,10 @@ export class Batch {
 
 		var next_batch = /** @type {Batch | null} */ (/** @type {unknown} */ (current_batch));
 
-		if (this.#pending === 0 && (this.#scheduled.length === 0 || next_batch !== null)) {
+		if (this.#pending === 0) {
 			this.#unlink();
 		}
 
-		// Edge case: During traversal new branches might create effects that run immediately and set state,
-		// causing an effect to be scheduled again. We need to traverse the current batch
-		// once more in that case - most of the time this will just clean up dirty branches.
-		// TODO I think we can delete this now since we re-iterate above
-		if (this.#scheduled.length > 0) {
-			if (next_batch !== null) {
-				for (const e of this.#scheduled) {
-					next_batch.#scheduled.push(e);
-				}
-
-				this.#scheduled = [];
-			} else {
-				next_batch = this;
-			}
-		}
-
 		if (next_batch !== null) {
 			old_values.clear();
 			next_batch.#process();

packages/svelte/src/internal/client/reactivity/deriveds.js

@@ -436,8 +436,9 @@ export function update_derived(derived) {
 			// to the previous batch, not the new one (which can already exist if an earlier
 			// effect wrote to a source). Not adding it to either can cause bugs, so we add it to both.
 			// See https://github.com/sveltejs/svelte/pull/18117 for more details.
-			previous_batch?.capture(derived, value, true);
+			// Order matters: current_batch first so it gets the correct previous value.
 			current_batch?.capture(derived, value, true);
+			previous_batch?.capture(derived, value, true, current_batch !== null);
 		} else {
 			derived.v = value;
 			derived.wv = increment_write_version();