The paper Fue muerto: Suppletion in Spanish Analytic Passives (p.96-112) analyses this very question.
The authors come to the conclusion that it is indeed suppletion:
We claim that there are three areas that provide more solid evidence for treating mat- and mor- as competing members of a list accessed by the same numeric index. The first is that past participles of mat- (matado, 'killed') and mor- (muerto, 'dead'), in the contexts of active perfects and eventive passives, have been in overlapping distribution in the history of the language and, for many modern speakers, demonstrate blocking effects typical of suppletive forms. In Old Spanish, the past participle of matar in both active (17a) and passive (17b) contexts surfaced as muerto (Corominas & Pascual, 1980-91).
a. Capitulo.xxxj. que muestra commo auariça ha muertos & condepnados grandes omnes (Libro de las donas, 1448)
'Chapter 31 shows how avarice has killed and condemned great men'
b. En el obispado de Nicaragua fueron prelados: primero, D. Fr...dominico, que fue muerto por los dos hermanos Contreras (Historia eclesiástica indiana, 1596)
'In the bishopric of Nicaragua there were prelates: first, D. Fr. Dominico, who was killed by the Contreras brothers'
In modern Spanish, muerto does not productively appear in active perfects (18a), but continues to be, for the vast majority of speakers, the only way of passivizing matar (18b).
a. He muerto una liebre3 (Archaic entry from the DRAE)
I have died a hare
'I have killed a hare'
b. Un ciclista fue {muerto/%matado4} en la calle
A cyclist was died/killed in the street
'A cyclist was killed in the street'
4. We use % to indicate variable acceptability. Some speakers reject matado completely while others find it marginally acceptable. Others may accept both muerto and matado with no intuitive difference in meaning. Additionally, some speakers have communicated that there is a difference in meaning between the two but have trouble saying exactly what that is. We feel that in this latter case there might simply be a sense in which they mean different things because they are different forms and speakers sometimes attempt to link different meanings to different forms oven if they don't exist. There is much to investigate here, but we will limit ourselves to a description and analysis of the grammars that use muerto and not matado in passives as a first step in a more complete characterization of the phenomenon.
This claim is corroborated by the results of a comparison of active perfects and passives of matar in a corpus of modern newspaper Spanish. This is a corpus we compiled of modern journalistic prose from the past five years that included the Spanish news-paper El País, the Mexican newspaper El Universal and the Argentine newspaper La Nación. A summary of the data is provided in Table 1. The participle muerto is virtually excluded from active perfect constructions meaning 'have killed', yet it is overwhelmingly preferred in analytic passive constructions 'be killed.'
In order to include other genres besides journalistic prose, we consulted El Corpus de Referencia del Español Actual (CREA), which includes a wide variety of both written and oral genres. Data from the CREA corpus further support the idea that muerto is the most widely used form in analytic passives meaning 'be killed.'
a. Example of 'fue muerto' search (CREA) - 113 hits
del asesinato del mahatma [Gandhi, que fue muerto por un fanático] que lo acusaba de haber permitido la dimisión del territorio
'Gandhi, who was killed by a fanatic'
(ORAL - Informe semanal 3/11/84 - TVE-1, CREA search 4/20/2013)
b. Example of 'fue matado' search (CREA) - 8 hits
Gaudí, cuyo estilo sinuoso, sensual y revolucionario puede ser visto por toda Barcelona (él la convirtió en su ciudad) fue matado por un tranvía ...
'Gaudí... was killed by a streetcar'
(Carlos Fuentes, El espejo enterrado, CREA search 4/20/2013)
The data in (18) - (19) and Table 1 indicate that the connection between mat- and mor- extends beyond the causative alternation and constitutes an empirical reality in the realm of participles from both a diachronic and a synchronic perspective. Diachronically, it appears that the syntactic environment of past participles (active perfects and eventive passives) commonly blocked the presence of the vocabulary item mat- while synchronically it is the combination of past participle and unaccusative syntax that blocks mat-.
...
5. Conclusion
In this paper we have tackled the controversial issue of whether suppletion, or root allomorphy, exists in human language. We have shown based on a grammatically-conditioned alternation between the words typically associated with the meanings KILL (matar) and DIE (morir) that suppletion is an empirical reality in Spanish and we provided an analysis of it using new insights by Harley (2014). This particular analysis, if on the right track, adds support to a broad view of grammar in which words are not atomic units that are opaque to syntactic operations but rather are products of syntactic operations that must be interpreted at the interfaces, which is encompassed by many frameworks such as Borer's (2013) exo-skeletal model, DM, and Nanosyntax (Caha, 2009; Fábregas, 2014; Starke, 2009). Beyond this, the details discussed in Section 3 add credence to the idea that roots are not individuated phono-logically or semantically in the syntax, which is a point of debate among practitioners of these various models.
