Internet disruptions abroad: what can your eSIM actually protect?

The findings, before buying another eSIM

3 scheduled notices and 0 continuity cases are counted separately. The incident count belongs to this corpus; it does not measure annual outage frequency.

Five destinations, not five countries ranked by risk

Thailand, Vietnam, Indonesia / Bali, Japan, Nepal

The selection is editorial: documented examples, usefulness for trip planning and presence in the supplier’s technical catalogue. It is neither a top-five tourism ranking nor a random sample. An incident absent from this page is not proof of a fault-free network. The corpus is non-exhaustive and does not compare annual availability between countries.

Where the buyer lives does not by itself determine the network used abroad. Regionalisation changes destinations and explanations; it does not manufacture new measurements. Historical facts about the same country retain their date and limitations across editions.

An eSIM provides access to an infrastructure chain

The downloadable profile replaces a SIM’s physical medium, not the towers or their power supply. If radio access fails, a data exit in another country does not restore that access. If only one application fails, maps and other messaging services may still work. Diagnosis needs to identify the failed dependency.

Not every “outage” means the same thing

Case register and nature of the evidence

Intervals are described only as far as the source permits. Traffic windows, blocking periods and notices are not added to produce “hours without Internet”. Each record states its scope; provider statements and traffic observations are different kinds of evidence.

The five destinations, case by case

What the eSIM Access technical catalogue reveals

The snapshot was collected on 13 September 2026 (2026-09-13T11:05:12.815Z). The full catalogue contained 3046 references. This technical panel retains 77 unique codes whose location equals the destination ISO and dataType = 1. Regional bundles and daily or other product classes are excluded from this panel.

The coverage column includes national products and products listing the country in a region. It is not added into a unique total: the same code can cover several destinations. Configurations are grouped by country, ipExport value and a sorted set of MNO names/network types. Different durations and allowances can share one configuration.

Advertised networks, without a switching promise

In this panel, 15 references list one MNO name and 62 list more than one. Counting names does not establish priority, simultaneous coverage, manual selection or automatic switching during a fault. No brand is assigned a backup capability without testing the purchased profile.

A commercial name may represent a merged brand or share infrastructure with another name. Retaining two literal names in the data does not imply two separate radio networks. The list is published as a dated supplier declaration, not an independent certification.

IP exit: a useful clue, not a route measurement

ipExport is retained literally. It may identify a country or a broad region such as Europe. It cannot reconstruct every hop, identify the profile APN, confirm an IMSI or establish the ASN the buyer will use. It also does not identify the country of a speedtest server.

The speedtest server is the test’s destination, not necessarily the connection’s exit. Publishing an observed route requires recording the public IP, ASN and multiple geolocation sources alongside the connected MNO, time, device and test. Where possible, traceroute adds context; hidden hops or conflicting geolocation require keeping the uncertainty.

An exit outside the destination may change latency or the filtering encountered by a connection, but it does not guarantee greater speed or access to blocked services. This is a technical hypothesis for future tests, not a result observed during the incidents in this corpus.

Why a 5G label is not enough

Advertised technology describes a contractual or network possibility, not the state of a particular connection. The handset, bands, area, roaming and plan conditions all matter. If speed says 3G/4G while locationNetworkList advertises 5G, both fields remain in the data; the contradiction is not resolved by inventing a capability.

A quality comparison requires the same device, location and time window, with provider order alternated. Observed 4G/5G connection, latency, jitter, loss, download and upload need to be recorded alongside practical tasks. One fast speedtest does not establish availability during a fault, and this study publishes no provider podium.

A backup can also expire or require Internet

An alternative that is not yet installed may depend on the failed connection to download its profile. Before travelling, confirm receipt, compatibility, instructions and activation mechanism. eSIM Access documentation distinguishes installation-triggered activation from first connection at the destination; the actual code for each reference is retained without generalising to every brand.

Validity is not assumed to pause during an outage, and data or days are not assumed to be restored automatically. Hotspot, refund or FUP rules are not inferred from an MNO list. These conditions require the exact product terms and, for real-world validation, a record of purchase, installation, connection and expiry.

eSIM Access: Plan Activation Mechanisms & Product Classification

How to prepare a backup that changes a dependency

1. Save maps, tickets, the address, reservation number and accommodation contact offline. A documentary backup remains useful even when every reachable network fails.
2. Identify the MNO behind the main and backup plans, not just the seller’s logo. Ask whether manual selection of another network is allowed and retain the exact answer.
3. Install over a stable connection only when the activation rules allow it without losing days. Keep instructions without publishing QR codes, ICCIDs or credentials.
4. During a fault, compare at least two unrelated services. If maps work but one application does not, avoid treating the problem as the loss of all data.
5. Check allowance and expiry, briefly toggle airplane mode, and inspect the registered network and instructed APN. Do not delete the profile as a first step: reinstallation may require authorisation or connectivity.
6. Use Wi-Fi or another available network where appropriate, without assuming every Wi-Fi path is independent. Follow official notices and avoid testing critical services or generating heavy traffic during an emergency.

For Asia: treat Bali’s Nyepi notices as a planning constraint, not as a measured test of every tourist profile. In Nepal, national coverage is not a guarantee along a trekking route. In Japan, the Noto example concerns affected areas, not an outage of Tokyo or every tourist city. Save maps and accommodation details before moving between countries and check the exact regional plan coverage.

What field tests contribute and what is still missing

Existing field measurements describe a connection at its test location and time. They are not retrospectively converted into measurements during a historical fault. Future Thailand tests can couple registered network, IP exit and speedtest, but cannot establish what happened in Canada or Spain.

Browse published field measurements

Reproducible protocol and counting rules

1. Fix the period and five destinations before counting. Retain primary sources: operators, regulators, technical registries, traffic observatories or blocking measurements. Exclude unverified rumours.
2. Create one episode/country record with date, scope, category, source and limitation. One blackout affecting Spain and Portugal creates two country records, not two independent causes.
3. Distinguish traffic observation, blocking measurement, operator statement, scheduled notice and continuity. A traffic reduction against the previous week is not the percentage of disconnected users.
4. Query /open/package/list using the existing integrated client without buying profiles. Export the selected technical fields; exclude wholesale costs, secrets and subscriber data.
5. Filter national references by location exactly equal to the ISO and dataType = 1. Deduplicate by packageCode. Group configurations by country + literal ipExport + sorted MNO/technology.
6. Keep the original catalogue timestamp separate from the editorial review date. Publish JSON and linked sources. Do not fill missing fields with assumptions.

Panel formula: reference count = cardinality of unique packageCode values after filtering. Configuration count = sum of within-country groups. Incident count = records not classified as notice or continuity. No denominator of observed hours is calculated, so there is no availability rate.

What this edition cannot establish

• It does not identify the country with the most outages or an annual traveller-risk rate.
• It does not rank providers by resilience. One wholesaler’s catalogue does not represent the whole market.
• It does not verify automatic switching, MNO priority, hotspot, APN, IMSI, ASN or the profile’s actual route.
• It does not assign an ASN’s traffic reduction or duration to each user. Networks and devices do not have the same observation scope.
• It does not assume every blocked service becomes accessible through a foreign IP, and does not recommend evading local rules.
• It does not assess emergency calls, SMS or rescue availability on a data-only eSIM. It is not a real-time safety tool.

Commercial transparency

Drift Asia sells eSIMs and uses eSIM Access as a supplier. Emmanuel Cerisola runs this activity and publication. The catalogue is a supplier declaration, not an independent audit of its network agreements. Drift Asia is neither certified nor given an advantage over other sellers.

No outages were induced, no profiles were bought for this edition, and no support or refund requests were sent. Customer tests will be published separately when evidence exists. No synthetic score or index is added to these data.

Frequently asked questions

Primary sources and traceability

1. Q2 2025 Internet disruption summary · Cloudflare Radar
2. Q3 2024 Internet disruption summary · Cloudflare Radar
3. Q4 2025 Internet disruption summary · Cloudflare Radar
4. Noto earthquake: mobile service recovery notice, updated 27 June 2024 · NTT DOCOMO
5. Japan's Telecommunications Carriers Join Forces to Strengthen Disaster Response · KDDI / NTT / SoftBank / Rakuten
6. Vietnam blocked Telegram · OONI
7. Seruan Bersama - Nyepi Çaka 1946 (2024) · Gouvernement provincial de Bali / FKUB
8. Seruan Bersama Pelaksanaan Hari Suci Nyepi, Pemprov Bali Himbau Jaga Ketertiban · Bali Media Center / Pemprov Bali
9. Seruan Bersama - Hari Suci Nyepi Çaka 1948 / 2026 · Badan Kesbangpol / Pemprov Bali

Editorial review: 14 September 2026. Source pages may be updated; the episode date remains in each record. The catalogue retains its 13 September timestamp and is not presented as a publication-day collection.

Data to cite and download

Suggested citation: Drift Asia, “Internet disruptions abroad: what can your eSIM actually protect?”, Asia edition, 14 September 2026. Retain the catalogue date, five destinations and distinction between incidents, notices and continuity.

https://driftasia.com/data-index/internet-disruptions-esim/

Original counts and analysis may be cited with attribution and a link. Third-party measurements and statements retain their own attribution; this page does not relicense their content. The technical JSON excludes wholesale costs, credentials and subscriber profiles.

Related studies and resources

• All published studies
• Press and data
• Overview and dated publications
• Field measurements