Flux metaquery causes "invalid memory address or nil pointer dereference"

If I run this query in cloud, I get a panic:

import "influxdata/influxdb/schema"
import "strings"

bucket = "apps"

schema.tagValues(bucket: bucket, tag: "_measurement")
  |> map(fn: (r) => {
    meas = r._value
    fields = schema.fieldKeys(bucket: bucket, predicate: (r) => r._measurement == meas) |> findColumn(column: "_value", fn: (key) => true)
    joinedFields = strings.joinStr(arr: fields, v: ", ")

    tags = schema.tagKeys(bucket: bucket, predicate: (r) => r._measurement == meas) 
      |> filter(fn: (r) => r._value != "_start" and r._value != "_stop" and r._value != "_measurement" and r._value != "_field")
      |> findColumn(column: "_value", fn: (key) => true)
    joinedTags = strings.joinStr(arr: tags, v: ", ")
    return {measurement: r._value, fields: joinedFields, tags: joinedTags}
  })
  |> group(columns: ["measurement"])

I get this panic: runtime error: invalid memory address or nil pointer dereference

It seems to work for some buckets, but not for others.

Start by reproducing the supplied Flux query in a bucket where it panics and comparing it with a bucket where it succeeds. Focus on the nested schema.tagValues, schema.fieldKeys, schema.tagKeys, findColumn, and strings.joinStr calls, then inspect the resulting panic details. Done means the query no longer produces an invalid memory address or nil pointer dereference for affected buckets.