diff --git a/src/gallium/drivers/zink/zink_context.c b/src/gallium/drivers/zink/zink_context.c
index 144ce2e7ea0..6518c3b271c 100644
--- a/src/gallium/drivers/zink/zink_context.c
+++ b/src/gallium/drivers/zink/zink_context.c
@@ -3172,6 +3172,7 @@ static void
 flush_batch(struct zink_context *ctx, bool sync)
 {
    struct zink_batch *batch = &ctx->batch;
+   assert(!ctx->unordered_blitting);
    if (ctx->clears_enabled)
       /* start rp to do all the clears */
       zink_batch_rp(ctx);
@@ -3502,7 +3503,7 @@ zink_set_framebuffer_state(struct pipe_context *pctx,
    ctx->gfx_pipeline_state.rast_samples = rast_samples;
 
    /* this is an ideal time to oom flush since it won't split a renderpass */
-   if (ctx->oom_flush)
+   if (ctx->oom_flush && !ctx->unordered_blitting)
       flush_batch(ctx, false);
    else
       update_layered_rendering_state(ctx);
diff --git a/src/gallium/drivers/zink/zink_draw.cpp b/src/gallium/drivers/zink/zink_draw.cpp
index f4efa6389a0..7d39a94c7f5 100644
--- a/src/gallium/drivers/zink/zink_draw.cpp
+++ b/src/gallium/drivers/zink/zink_draw.cpp
@@ -906,7 +906,7 @@ zink_draw(struct pipe_context *pctx,
    batch->last_was_compute = false;
    ctx->batch.work_count = work_count;
    /* flush if there's >100k draws */
-   if (unlikely(work_count >= 30000) || ctx->oom_flush)
+   if (!ctx->unordered_blitting && (unlikely(work_count >= 30000) || ctx->oom_flush))
       pctx->flush(pctx, NULL, 0);
 }
 
@@ -1079,7 +1079,7 @@ zink_launch_grid(struct pipe_context *pctx, const struct pipe_grid_info *info)
    batch->has_work = true;
    batch->last_was_compute = true;
    /* flush if there's >100k computes */
-   if (unlikely(ctx->batch.work_count >= 30000) || ctx->oom_flush)
+   if (!ctx->unordered_blitting && (unlikely(ctx->batch.work_count >= 30000) || ctx->oom_flush))
       pctx->flush(pctx, NULL, 0);
 }