memory allocation limit reached: limit 10000000 bytes, allocated: xxxxxxxx, wanted: xxxxx

__Steps to reproduce:__
1. Happens on any dashboard that queries a lot of data with a simple query
e.g.
`from(bucket: "PLC_Router_Data")
|> range(start: v.timeRangeStart, stop: v.timeRangeStop)
|> filter(fn: (r) => r["_measurement"] == "data")
|> filter(fn: (r) => r["host"] == v.my_PLC_source)
|> filter(fn: (r) => r["_field"] =~ /ress/ and r["_field"] !~ /Application/)
|> group(columns: ["_field"], mode:"by")
|> toFloat()
|> sort(columns: ["_time"], desc: false)`

__Expected behavior:__
I expect that performing the same query on the same dataset for the same duration either results in:
1) The query *always fails* if we are genuinely doing something that consumes massive amounts of memory (it succeeds 50% of the time)
OR
2) The query *always* succeeds if it's not

__Actual behavior:__
Sometimes you get data

Sometimes you get the failure

__Environment info:__
* InfluxDB 2.0 Cloud