Flaky `test_scatter_compute_lose`

Might be a recent regression:

```
__________________________ test_scatter_compute_lose ___________________________
c =
s =
a =
b =
@gen_cluster(client=True)
asyncdeftest_scatter_compute_lose(c, s, a, b):
[x] = await c.scatter([[1, 2, 3, 4]], workers=a.address)
y = c.submit(inc, 1, workers=b.address)
z = c.submit(slowadd, x, y, delay=0.2)
await asyncio.sleep(0.1)
await a.close()
with pytest.raises(CancelledError):
> await wait(z)
E Failed: DID NOT RAISE
distributed/tests/test_client.py:3942: Failed
----------------------------- Captured stderr call -----------------------------
2022-07-15 17:50:19,265 - distributed.worker - WARNING - Compute Failed
Key: slowadd-12ea486410272c2ea75c5d36876e43d0
Function: slowadd
args: ([1, 2, 3, 4], 2)
kwargs: {'delay': 0.2}
Exception: 'TypeError(\'can only concatenate list (not "int") to list\')'
```
https://github.com/dask/distributed/runs/7362409614?check_suite_focus=true#step:11:1271