Once dz and z had merged, a scribe writing Slavonic had two letters available for a single sound and no way to decide between them by ear. The response, repeated across the tradition, was to fix the choice by word rather than by sound: certain words are spelled with one letter and certain words with the other, and a scribe learns the list.
This is the single most characteristic feature of Church Slavonic orthography, and dzelo is only its first instance. The same problem and the same solution recur with on and omega, both writing o; with izhe and i, both writing i; with fert and fita, both writing f; and eventually with yat, which stopped being distinct in speech long before it stopped being written. In each case the pair survives because the church books preserve the distinction, and the distinction survives because it has become a matter of correct spelling.
The effect is that literacy in Church Slavonic is not simply the ability to sound out letters. It requires knowing a body of conventions no pronunciation will teach, which raises the cost of writing correctly and makes orthography a mark of schooling. That is not an accident of neglect; it is what happens to any orthography that outlives the sound system it was designed for, and it is the same pressure that later produced spelling reforms in Russia, Bulgaria and Serbia.
Reformers in all three countries began by attacking exactly these pairs, and the arguments they met were not about phonetics.