Tony Beamish

Two Labour Appeal Court judgments delivered in September 2026 contain unusual apologies. In each, the judge who ultimately wrote the judgment explained that another judge had originally been assigned to do so.

Both matters were heard on 24 February 2026. The judgments followed on 7 and 10 September, 195 and 198 calendar days later. Their explanations raise questions about responsibility for delayed judgments and about what the public can establish from the available records.

Two judgments and an unanswered question

The same three acting judges of appeal sat in both matters: Bantubonke Tokota, Graham Moshoana and Colleen Collis.

In Itec As (Pty) Ltd v Ngwane NO and Others, delivered on 7 September, Moshoana apologised for the delay and disclosed: “I was not the allocated scribe of this judgment.” He explained that the original writer had been unable to produce it timeously and that he had volunteered, as directed, to write it. Tokota and Collis concurred.

Three days later, in Ndaba v South African Social Security Agency and Others, Tokota explained in paragraph 45 that the matter had initially been assigned to another judge. He said that he and Moshoana had decided to write the two judgments in the interests of justice and apologised for the delay. Moshoana and Collis concurred.

Tokota said the original writer had offered repeated assurances that the judgments would be forthcoming, but that he and Moshoana had received no explanation for the delay. He also acknowledged their responsibility to keep reminding their colleague.

Read together, the composition of the bench and the two judges’ accounts strongly suggest that Collis was the original designated writer. Neither judgment names her in that capacity. That is an inference from the published judgments, rather than an express finding, and the reasons for the original writer’s delay are not established by those accounts.

The distinction matters. Lengthy delays deserve scrutiny, but a judgment’s delivery date alone does not establish why it was delayed or which judge was responsible for each part of the interval. The writing allocation, subsequent directions and any explanation from the judges concerned are necessary to complete that picture.

What the judicial standards require

Paragraph 5.2.6 of the judicial norms and standards issued in 2014 requires every effort to deliver judgments within three months of the last hearing, save in exceptional cases where that is not possible. The benchmark runs from the last hearing. It does not create a general three-month period starting whenever a litigant asks for reasons.

Article 10 of the Code of Judicial Conduct also requires diligence and prompt judgments. Article 10(2) provides for delivery within the term in which the hearing was completed, with specified exceptions allowing delivery during the next term. Assessing compliance requires the relevant circumstances; an interval above three months is not, by itself, a finding of judicial misconduct.

The concern is longstanding. In Pharmaceutical Society of South Africa and Others v Minister of Health and Another; New Clicks South Africa (Pty) Limited v Tshabalala-Msimang NO and Another, decided in December 2004, Harms JA criticised the delay in deciding applications for leave to appeal. He emphasised, by quoting an earlier decision, that judicial delay can obstruct the administration of justice.

What the public records can show

In May 2026 I submitted a request under the Promotion of Access to Information Act (PAIA) to the Office of the Chief Justice (OCJ) for records concerning judgments in the Gauteng Division.

The response I received gave three grounds for refusing or redirecting the request: insufficient identification of the requester in that my request was not accompanied by a copy of my South African identity card; the work that would be required to search and collate records over more than four years; and the asserted public availability of the information in court rolls and reported judgments.

Those explanations raise a practical question: could the relevant information be supplied from existing monitoring records, without reconstructing every court file? The OCJ’s role in supporting judicial performance monitoring was described in the Justice Minister’s July 2024 budget speech. That does not prove that the particular dataset I requested exists in the requested form.

Section 45(b) of PAIA permits refusal where processing would substantially and unreasonably divert a public body’s resources. The Act also addresses assistance with defective requests. Whether the refusal in my case was justified must be assessed against the request, the response and the relevant records.

There is already an official source that should form part of the enquiry. The judiciary publishes reserved judgment reports listing outstanding matters. The report for the beginning of the fourth term of 2025 identifies matters outstanding for more than six months. That reporting threshold is distinct from the three-month norm. It also measures a backlog at a point in time, rather than the time taken to deliver the judgments in my sample.

I began examining published judgments on the South African Legal Information Institute (SAFLII) database. That work can illuminate individual delays, but it cannot by itself supply a complete account of a judge’s workload or output.

How the figures should be read

The figures below describe the entries collected in my spreadsheets. They are provisional sample statistics, not an official assessment or a ranking of judges. Publication on SAFLII, the availability of hearing dates and the selection of entries all affect what is counted. Orders, oral rulings and matters not published in the database may be absent.

For this comparison, elapsed time is calculated in calendar days between the reserved dates and judgment dates recorded. Entries lacking either date are excluded from the average; same-day entries are included as zero. Where a judgment records several hearing dates, the last hearing date was used by me.

Each judgment with both dates receives the same weight. Averaging annual averages gives equal influence to years with very different numbers of judgments, so it is not a substitute for calculating the mean across individual entries. These samples also cover different periods and types of case; the averages cannot fairly be read as a measure of relative effort or judicial quality.

Judge Bantubonke Tokota

Tokota acted in the High Court and Labour Court between 2004 and 2017 and became a permanent Eastern Cape High Court judge in October 2017. His official biography also records acting appointments in the Labour Appeal Court in 2022 and the Supreme Court of Appeal in late 2023.

My Tokota spreadsheet contains 78 judgment entries. Seventy-seven have both dates needed for the calculation. Their combined elapsed time is 3,078 days, giving a provisional average of 40.0 days. This describes the sample collected, rather than every judgment he has delivered since 2004.

Judge Graham Moshoana

The Moshoana spreadsheet supplied for this article contains 189 entries. Of these, 188 have both dates needed for the calculation. Using the dates as entered, including 23 same-day entries, the provisional average is 11.8 days. SAFLII suggests that Moshoana has authored 455 judgments but only the first 200 can presently be accessed on its database. The figures here will be updated once the SAFLII database access problem has been resolved.

Judge Colleen Collis

Collis’s permanent appointment to the Gauteng Division of the High Court took effect on 1 January 2018.

The recorded delays in Collis’s judgment are substantially longer in the 2024–2026 than in the earlier years shown. That pattern warrants investigation. It does not establish why the delays occurred, and are not presented as a complete measure of performance.

Year	Entries collected	Entries with both dates	Mean calendar days
2018	13	12	56.4
2019	33	26	121.1
2020	27	26	93.0
2021	21	21	108.0
2022	14	13	117.5
2023	10	10	112.1
2024	18	18	260.3
2025	17	17	229.6
2026	10	10	245.8

In the ten 2026 entries, the recorded intervals range from eight to 446 days and average 245.8 days. Five are close to or beyond a year: 356, 363, 372, 408 and 446 days. The longest is Afrinergy Holdings v Commissioner, South African Revenue Services; my workbook records a hearing on 13 November 2024 and delivery on 2 February 2026.

What would improve accountability

A more useful public record would show when a matter was last heard, when judgment was reserved, who was assigned to write it, whether that assignment changed, and when the order and reasons were delivered. Where a judgment remains outstanding, the record should explain any revised delivery date and the steps being taken to resolve it.

Better reporting would help distinguish an unexplained failure to deliver from a delay with a documented reason. It would also allow the public to scrutinise both individual responsibility and the support and supervision provided by court leadership.

Technology may assist legal research and the preparation of judgments, but its ability to reduce these delays still needs to be demonstrated. It cannot replace a judge’s responsibility for the decision or the court’s responsibility to manage outstanding work.

The apologies in Itec and Ndaba are a starting point. The remaining questions concern what happened before the judgments were reassigned, why the parties waited as long as they did, and how similar delays will be prevented.

Disclosure: I am also awaiting reasons from a Gauteng judge who is not one of the judges discussed in this article. 643 days, and counting.