`pytest --dist=loadgroup` hangs if a crashed worker is restarted - variant 2

[radoering]

This issue is similar to #1323 (regarding the issue that can be observed from the outside), but since it is triggered by a slightly different setup, the root cause is different and it requires a different fix, I created a new issue.

When running the following test file

def test_1():
    import time
    time.sleep(5)
    assert True

def test_2():
    assert True

with

pytest -n1 --dist=loadgroup -o faulthandler_timeout=1 -o faulthandler_exit_on_timeout=true testing/test_timeout.py

test execution hangs with

replacing crashed worker gw0
2 workers [2 items]

The issue is that only one item is assigned to the new worker in

pytest-xdist/src/xdist/scheduler/loadscope.py

Lines 352 to 355 in 8fed345

	if self.collection is not None:
	for node in self.nodes:
	self._reschedule(node)
	return

This item is removed from the queue in

pytest-xdist/src/xdist/remote.py

Line 204 in 8fed345

self.nextitem_index = self.torun.get()

Then, in

pytest-xdist/src/xdist/remote.py

Line 214 in 8fed345

self.nextitem_index = self.torun.get()

the worker waits for a second item.

In contrast, at first start (when self.collection is still None), two items are added to the queue via

pytest-xdist/src/xdist/scheduler/loadscope.py

Lines 396 to 402 in 8fed345

	# Assign initial workload
	for node in self.nodes:
	self._assign_work_unit(node)
	# Ensure nodes start with at least two work units if possible (#277)
	for node in self.nodes:
	self._reschedule(node)