Elasticsearch version (bin/elasticsearch --version): 6.7, 7.9.2, 7.16.2

Plugins installed: [none]

JVM version (java -version): 1.8

OS version (uname -a if on a Unix-like system): docker, ubuntu 18.04

I experience a problem with the completion suggester in Elasticsearch 7.9.2 (also present in 6.7 and 7.16.2)

If I have more than one suggest input with the same value for different contexts it will only find a completion for the first one, not the second.

The suggest Part of the Document looks like this.

"SUGGEST": [ { "input": [ "Inzidenz", "Notbremse", "Kraft", "Mallorca" ], "contexts": { "FIELD": [ "DESKT", "DESKT_ABST" ] } }, { "input": [ "Pandemie", "Notbremse", "Kraft" ], "contexts": { "FIELD": [ "TIT", "TITSPLABST", "TITSP", "TITTXTSP", "TIT_RHTI" ] } } ]

Both contein the input "Notbremse" But if I search for "Notbre" with the contexts FIELD "TITSPLABST" I get no results. But if I search for "Notbre" with the contexts FIELD "DESKT" i get the expected result. It depends on the order of the input fields in the document. if i change the order that the input field for "TITSPLABST" is before the other then it works.

Steps to reproduce:

Please include a minimal but complete recreation of the problem,
including (e.g.) index creation, mappings, settings, query etc. The easier
you make for us to reproduce it, the more likely that somebody will take the
time to look at it.

1. create Index

`PUT /suggest_test?pretty
{
"settings" : {
"index.mapping.nested_fields.limit" : 60,

	"number_of_replicas" : 1,
	"number_of_shards" : 6,
	
	"index" : {
		"analysis" : {
			"char_filter" : {
				"umlaut_filter" : {
					"type" : "mapping",
					"mappings" : ["\u00c4=>ae", "\u00e4=>ae", "\u00d6=>oe", "\u00f6=>oe", "\u00dc=>ue", "\u00fc=>ue", "\u00df=>ss", "\u00c6=>Ae", "\u00e6=>ae", "\u00d8=>Oe", "\u00f8=>oe"]
				},
				"underscore_pattern" : {
					"type" : "pattern_replace",
					"pattern" : "\\_",
					"replacement" : ""
				}
			},
			"tokenizer" : {
				"sd_tokenizer" : {
					"type" : "pattern",
					"pattern" : "[^\\p{N}\\p{L}#]+",
					"group" : "-1"
				}
			},
			"filter" : {
				"prefixFilter" : {
			               "type" : "pattern_capture",
			               "preserve_original" : true,
			               "patterns" : [
			                  "([a-zA-Z]+[^#])$"
			               ]
			            }     
			},
			"analyzer" : {
				 "default": {
					"type" : "custom",
					"tokenizer" : "sd_tokenizer",
					"char_filter" : ["umlaut_filter", "underscore_pattern"],
					"filter" : ["lowercase", "asciifolding"]
				},
				"prefixAnalyzer": {
					"tokenizer" : "sd_tokenizer",
					"char_filter" : ["umlaut_filter", "underscore_pattern"],
					"filter" : ["prefixFilter","lowercase", "asciifolding"]
				},
				"same_analyzer": {
					"type" : "custom",
					"tokenizer" : "keyword",
					"char_filter" : ["umlaut_filter", "underscore_pattern"],
					"filter" : ["lowercase", "asciifolding"]
				},
				
				"online_rechte_analyzer": {
					"type" : "custom",
					"tokenizer" : "sd_tokenizer",
					"filter": ["lowercase"]
				}
			}
		}
	}
},
"mappings": {
		"_source" : {
			"enabled" : true
		},
		"dynamic" : "strict",
		"properties" : {
			"AK" : {
				"properties" : {
					"TITLE" : {
						"type" : "text"
					},
					"TEXT" : {
						"type" : "text"
					}
			}
		},
		"SUGGEST": {
				"type": "completion",
				"analyzer": "default",
				"preserve_separators": true,
				"preserve_position_increments": true,
				"max_input_length": 25,
				"contexts": [
					{
						"name": "FIELD",
						"type": "category"
					}
				]
		}  
	}
}

}`

2. Put Document into index

POST /suggest_test/_doc?refresh=true { "AK": [ { "TITLE": "Die Notbremse tritt in Kraft", "TEXT": [ "Inzidenz von 99,9, ab 100 soll die vereinbarte 'Notbremse' in Kraft treten.\n\n(O-Ton) Angela Merkel: \"Werden leider von der Notbremse Gebrauch machen muessen\" \n(O-Ton) Michael Kretschmer: Schnell umsteuern \n(O-Ton) Malu Dreyer: Oeffnung Aussengastronomie bei Kontaktbeeschraenkungen \n(O-Ton) Manuela Schwesig: Contro Reisen nach Mallorca, aber nicht im eigenen Bundesland \n(O-Ton) Angela Merkel: \"Devise lautet: Impfen, impfen, impfen \n(O-Ton) Ulrich Weigelt: Die Menschen moechten beim Hausarzt geimpft werden \n(O-Ton) Manuela Schwesig: (Impfstoff Sputnik V) Impfstoff muss sicher sein, Herkunft darf nicht entscheidend sein" ] } ], "SUGGEST": [ { "input": [ "Inzidenz", "Notbremse", "Kraft", "Mallorca" ], "contexts": { "FIELD": [ "DESKT", "DESKT_ABST" ] } }, { "input": [ "Pandemie", "Notbremse", "Kraft" ], "contexts": { "FIELD": [ "TIT", "TITSPLABST", "TITSP", "TITTXTSP", "TIT_RHTI" ] } } ] }

3. Search for completion of "Notbrem" in contexts FIELD TITSPLABST (TITSPLABST is just a name from the Java context)
   This search produces no result

POST suggest_test/_search?pretty { "suggest": { "hfdb-suggest" : { "prefix" : "Notbrem", "completion" : { "field" : "SUGGEST" ,"skip_duplicates":false,"contexts":{ "FIELD":[{"context":"TITSPLABST","boost":1}] } } } } }

4. Now search for "Notbrem" in contexts FFIELD DESKT
   This search produces a result

POST suggest_test/_search?pretty { "suggest": { "hfdb-suggest" : { "prefix" : "Notbrem", "completion" : { "field" : "SUGGEST" ,"skip_duplicates":false,"contexts":{ "FIELD":[{"context":"DESKT","boost":1}] } } } } }

I expect a result in both queries, but only got one in the second.

Provide logs (if relevant):