Workaround to avoid false-positive version warning (#6656)

* implement workaround to avoid false-positive version warning

* update changelog

CHANGELOG.unreleased.md

@@ -35,9 +35,12 @@ For upgrade instructions, please check the [migration guide](MIGRATIONS.released
 ### Fixed
 - Fixed a bug in the dataset import view, where the layer name text field would lose focus after each key press. [#6615](https://github.com/scalableminds/webknossos/pull/6615)
 - Fixed importing NGFF Zarr datasets with non-scale transforms. [#6621](https://github.com/scalableminds/webknossos/pull/6621)
+- Fixed a regression in NGFF Zarr import for datasets with no channel axis. [#6636](https://github.com/scalableminds/webknossos/pull/6636)
+- Fixed broken creation of tasks using base NMLs. [#6634](https://github.com/scalableminds/webknossos/pull/6634)
 - Fixed a regression in NGFF Zarr import for datasets with no channel axis. [#6636](https://github.com/scalableminds/webknossos/pull/6636
 - Fixed broken creation of tasks using base NMLs. [#6634](https://github.com/scalableminds/webknossos/pull/6634)
 - Fixed that the precomputation of meshes didn't take the active mapping into account. [#6651](https://github.com/scalableminds/webknossos/pull/6651)
+- Fixed false-positive warning about an outdated annotation version. [#6656](https://github.com/scalableminds/webknossos/pull/6656)
 
 ### Removed
 

frontend/javascripts/oxalis/model/sagas/save_saga.ts

@@ -1147,6 +1147,25 @@ function* watchForSaveConflicts() {
         tracing.type,
       );
 
+      // There is a rare chance of a race condition happening
+      // which can result in an incorrect toast warning.
+      // It occurs if certain saga effects run in the following order:
+      // 1) a new version is pushed to the server (in sendRequestToServer)
+      // 2) the current server version is fetched (in this saga here)
+      // 3) the server version is compared with the client version (in this saga here)
+      // 4) only now is the local version number updated because the
+      //    sendRequestToServer saga was scheduled now.
+      // Since (4) is happening too late, the comparison in (3) will assume
+      // that client and server are out of sync.
+      // The chance of this happening is relatively low, but from time to time it
+      // happened. To mitigate this problem, we introduce a simple sleep here which
+      // means that the chance of (4) being scheduled after (2) should be close to
+      // 0 (hopefully).
+      // Note that a false-positive warning in this saga doesn't have serious side-effects
+      // (except for potentially confusing the user). This is why a more complex mutex- or
+      // semaphore-based solution to this problem was not chosen.
+      yield* call(sleep, 2000);
+
       // Read the tracing version again from the store, since the
       // old reference to tracing might be outdated now due to the
       // immutability.